0% found this document useful (0 votes)
29 views80 pages

Deep Learning in Data Analytics: Recent Techniques, Practices and Applications 1st Edition Debi Prasanna Acharjya

The document provides information about the book 'Deep Learning in Data Analytics: Recent Techniques, Practices and Applications', which discusses advanced computational theories and tools for analyzing large data sets using deep learning techniques. It highlights the importance of deep learning in extracting knowledge from voluminous data and addresses various applications in fields like healthcare and finance. The book is structured into four parts covering theoretical foundations, computing systems, deep learning algorithms, and practical applications, aiming to support researchers and practitioners in the field of data analytics.

Uploaded by

fmalswacha76
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
29 views80 pages

Deep Learning in Data Analytics: Recent Techniques, Practices and Applications 1st Edition Debi Prasanna Acharjya

The document provides information about the book 'Deep Learning in Data Analytics: Recent Techniques, Practices and Applications', which discusses advanced computational theories and tools for analyzing large data sets using deep learning techniques. It highlights the importance of deep learning in extracting knowledge from voluminous data and addresses various applications in fields like healthcare and finance. The book is structured into four parts covering theoretical foundations, computing systems, deep learning algorithms, and practical applications, aiming to support researchers and practitioners in the field of data analytics.

Uploaded by

fmalswacha76
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 80

Read Anytime Anywhere Easy Ebook Downloads at ebookmeta.

com

Deep Learning in Data Analytics: Recent


Techniques, Practices and Applications 1st Edition
Debi Prasanna Acharjya

https://ebookmeta.com/product/deep-learning-in-data-
analytics-recent-techniques-practices-and-applications-1st-
edition-debi-prasanna-acharjya/

OR CLICK HERE

DOWLOAD EBOOK

Visit and Get More Ebook Downloads Instantly at https://ebookmeta.com


Recommended digital products (PDF, EPUB, MOBI) that
you can download immediately if you are interested.

Social Big Data Analytics: Practices, Techniques, and


Applications Bilal Abu-Salih

https://ebookmeta.com/product/social-big-data-analytics-practices-
techniques-and-applications-bilal-abu-salih/

ebookmeta.com

Deep Learning for Biomedical Data Analysis: Techniques,


Approaches, and Applications 1st Edition Mourad Elloumi

https://ebookmeta.com/product/deep-learning-for-biomedical-data-
analysis-techniques-approaches-and-applications-1st-edition-mourad-
elloumi/
ebookmeta.com

Online Learning Analytics (Data Analytics Applications)


1st Edition Jay Liebowitz

https://ebookmeta.com/product/online-learning-analytics-data-
analytics-applications-1st-edition-jay-liebowitz/

ebookmeta.com

Minoru Yamasaki and the Fragility of Architecture 1st


Edition Paul Kidder

https://ebookmeta.com/product/minoru-yamasaki-and-the-fragility-of-
architecture-1st-edition-paul-kidder/

ebookmeta.com
The Legacy of Dell Hymes Ethnopoetics Narrative Inequality
and Voice 1st Edition Paul V Kroskrity Anthony K Webster

https://ebookmeta.com/product/the-legacy-of-dell-hymes-ethnopoetics-
narrative-inequality-and-voice-1st-edition-paul-v-kroskrity-anthony-k-
webster/
ebookmeta.com

A New Companion to Linguistic Anthropology 1st Edition


Alessandro Duranti Rachel George Robin Conley Riner

https://ebookmeta.com/product/a-new-companion-to-linguistic-
anthropology-1st-edition-alessandro-duranti-rachel-george-robin-
conley-riner/
ebookmeta.com

Cursed Shadows 1st Edition Kc Kean

https://ebookmeta.com/product/cursed-shadows-1st-edition-kc-kean/

ebookmeta.com

Dedicated to Mr Darcy A Pride and Prejudice Variation 1st


Edition Florence Gold A Lady

https://ebookmeta.com/product/dedicated-to-mr-darcy-a-pride-and-
prejudice-variation-1st-edition-florence-gold-a-lady/

ebookmeta.com

Indigenous Digital Life: The Practice and Politics of


Being Indigenous on Social Media Carlson

https://ebookmeta.com/product/indigenous-digital-life-the-practice-
and-politics-of-being-indigenous-on-social-media-carlson/

ebookmeta.com
Intermediate Maths for Chemists Chemistry Maths 2 2nd
edition J. E. Parker

https://ebookmeta.com/product/intermediate-maths-for-chemists-
chemistry-maths-2-2nd-edition-j-e-parker/

ebookmeta.com
Studies in Big Data 91

Debi Prasanna Acharjya


Anirban Mitra
Noor Zaman Editors

Deep Learning
in Data
Analytics
Recent Techniques, Practices and
Applications
Studies in Big Data

Volume 91

Series Editor
Janusz Kacprzyk, Polish Academy of Sciences, Warsaw, Poland
The series “Studies in Big Data” (SBD) publishes new developments and advances in
the various areas of Big Data- quickly and with a high quality. The intent is to cover the
theory, research, development, and applications of Big Data, as embedded in the fields
of engineering, computer science, physics, economics and life sciences. The books of
the series refer to the analysis and understanding of large, complex, and/or distributed
data sets generated from recent digital sources coming from sensors or other physical
instruments as well as simulations, crowd sourcing, social networks or other internet
transactions, such as emails or video click streams and other. The series contains
monographs, lecture notes and edited volumes in Big Data spanning the areas of
computational intelligence including neural networks, evolutionary computation,
soft computing, fuzzy systems, as well as artificial intelligence, data mining, modern
statistics and Operations research, as well as self-organizing systems. Of particular
value to both the contributors and the readership are the short publication timeframe
and the world-wide distribution, which enable both wide and rapid dissemination of
research output.
The books of this series are reviewed in a single blind peer review process.
Indexed by SCOPUS, SCIMAGO and zbMATH.
All books published in the series are submitted for consideration in Web of Science.

More information about this series at http://www.springer.com/series/11970


Debi Prasanna Acharjya · Anirban Mitra ·
Noor Zaman
Editors

Deep Learning in Data


Analytics
Recent Techniques, Practices
and Applications
Editors
Debi Prasanna Acharjya Anirban Mitra
School of Computing Science Department of Computer Science
and Engineering Engineering
VIT University Amity University Kolkata
Vellore, India Kolkata, India

Noor Zaman
School of Computer Science
and Engineering
Taylor’s University
Subang Jaya, Malaysia

ISSN 2197-6503 ISSN 2197-6511 (electronic)


Studies in Big Data
ISBN 978-3-030-75854-7 ISBN 978-3-030-75855-4 (eBook)
https://doi.org/10.1007/978-3-030-75855-4

© Springer Nature Switzerland AG 2022


This work is subject to copyright. All rights are reserved by the Publisher, whether the whole or part of
the material is concerned, specifically the rights of translation, reprinting, reuse of illustrations, recitation,
broadcasting, reproduction on microfilms or in any other physical way, and transmission or information
storage and retrieval, electronic adaptation, computer software, or by similar or dissimilar methodology
now known or hereafter developed.
The use of general descriptive names, registered names, trademarks, service marks, etc. in this publication
does not imply, even in the absence of a specific statement, that such names are exempt from the relevant
protective laws and regulations and therefore free for general use.
The publisher, the authors and the editors are safe to assume that the advice and information in this book
are believed to be true and accurate at the date of publication. Neither the publisher nor the authors or
the editors give a warranty, expressed or implied, with respect to the material contained herein or for any
errors or omissions that may have been made. The publisher remains neutral with regard to jurisdictional
claims in published maps and institutional affiliations.

This Springer imprint is published by the registered company Springer Nature Switzerland AG
The registered company address is: Gewerbestrasse 11, 6330 Cham, Switzerland
Dedicated to
My loving children Aditi and Aditya
D. P. Acharjya
My caring parents Amitava, Anushila and
wife Sudeepta
Anirban Mitra
My all family members
Noor Zaman
Preface

The amount of data collected across a wide variety of fields today far exceeds our
ability to reduce and analyze without automated analysis techniques. There is much
information hidden in the accumulated voluminous data. It is tough to obtain this
information. So, it is essential for a new generation of computational theories and
tools to help humans extract knowledge, self-learning, and rule generation from huge
data. Deep learning has evolved into an important and active research area because of
theoretical challenges associated with discovering intelligent solutions for the smart
analysis of huge data.
Deep learning is a form of machine learning that enables computers to learn from
experience and understand the world in terms of a hierarchy of concepts. Because the
computer gathers knowledge from experience, there is no need for a human–computer
operator to specify all the knowledge that the computer needs formally. The hier-
archy of concepts allows the computer to learn complicated concepts by building
them out of simpler ones; a graph of these hierarchies would be many layers deep.
The amount of data collected from various applications worldwide across various
fields is expected to double every two years. It has no utility unless these are analyzed
to get useful information. This necessitates the development of techniques that can be
used to facilitate machine-based rule generation and data analysis. The development
of powerful computers is a boon to implement these techniques leading to automated
systems. The transformation of data into knowledge is by no means an easy task for
high-performance large-scale data processing, including exploiting the parallelism
of current and upcoming computer architectures for data mining. Moreover, these
data may involve many challenges in terms of consistency or size, or dimension.
Advanced computing and learning techniques that include deep learning concepts
are fruitful in representing data and generating rules and knowledge. Further, these
models are very much fruitful for analysis also. The new challenges usually comprise,
sometimes even deteriorate, the performance, efficiency, and scalability of the dedi-
cated data-intensive computing systems. Deep learning approaches prevent loss of
information and hence enhance the performance of data analysis and learning tech-
niques. This brings up many research issues in the industry and research community
in capturing and accessing data effectively. Besides, fast processing while achieving
high performance and high throughput and storing it efficiently for future use is
vii
viii Preface

another issue. Further, programming for data analysis using deep learning concepts
is an important and challenging issue. Expressing data access requirements of appli-
cations and designing programming language abstractions to exploit parallelism is
an immediate need.
This book offers concepts and techniques of deep learning in a precise and clear
manner to the research community. In editing the book, we have tried our best
to provide frontier advances and applications in deep learning for data analysis,
the conceptual basis required to achieve in-depth knowledge in computer science
and information technology. This volume will help researchers interested in this
field to keep insight into different concepts and their importance for applications in
industry, research, and management fields. Further, this edited book aims to provide
computer science and information technology researchers with recent advances in
deep learning, which are required to achieve in-depth knowledge in computer science.
We believe our effort will make this edited volume interesting and attractive among
the student and research communities. This book is comprised of four sections. The
Part I is on the theoretical foundation of deep learning. The Part II is about computing
system and machine learning associated with deep learning techniques. Some of the
deep learning algorithms have been discussed in the Part III, and the Part IV focuses
on applications of deep learning techniques. This has been done to make the edited
book more flexible and stimulate further research interest in topics.
Theoretical foundation of deep learning consists of four chapters. The chapter
“A Study on Discrete Action Sequences Using Deep Emotional Intelligence” is on
discrete action sequences using deep emotional intelligence. Recognition of emotions
from human plays a vital role in our day to day life and is essential for social commu-
nication. Automatic emotion recognition is becoming a recent research focus on arti-
ficial intelligence. This chapter proposed a machine learning approach and discussed
with deep learning model to achieve emotional intelligence. Further, it provides a
brief study on achieving emotional intelligence with a DCNN Model. A novel noise
removal technique influenced by deep Convolutional autoencoders on mammograms
has been discussed in the chapter “A Novel Noise Removal Technique Influenced
by Deep Convolutional Autoencoders on Mammograms”. This chapter explains the
concepts and implementation of a deep learning algorithm for data analytics on struc-
tured and unstructured data concerning biomedical images such as CT image, MRI,
and X-ray to detect some deadly diseases.
The chapter “A High Security Framework Through Human Brain Using Algo
Mixture Model Deep Learning Algorithm” focuses on the security framework
through human brain using deep learning algorithm. This chapter aims to analyze
human brain patterns’ uniqueness and provide high-level security based on various
factors such as size, state of mind, and rate of waves. Chapter “Knowledge Framework
for Deep Learning: Congenital Heart Disease” is on knowledge framework for deep
learning with implementation on congenital heart disease. This chapter discusses
an automated strategy for analyzing, detecting, and diagnosing heart diseases with
initial valid parameters.
The Part II of this edited volume is computing system and machine learning asso-
ciated with deep learning techniques. This section consists of the next four chapters.
Preface ix

The chapter “Automatic Image Segmentation by Ranking-Based SVM in Convo-


lutional Neural Network on Diabetic Fundus Image” discusses automatic image
segmentation by ranking-based SVM in convolutional neural network on diabetic
fundus image. A two-stage deep learning approach has been proposed in this chapter
consisting of a deep ranking support vector machine used to define a consistent feature
label and add the output constrained layer of SVM to CNN for better acceptability
of the results.
Chapter “Deep Learning in Healthcare” focuses on issues of deep learning
in healthcare applications. In addition to diagnosis, deep learning can study the
disease’s progression, develop a personalized treatment plan, and overall patient
management. This chapter discusses the architecture and working of deep neural
networks and focuses on their application in detecting and treating various diseases
like cancer, diabetes, Alzheimer’s, and Parkinson’s disease. Application of machine
learning algorithms towards healthcare applications has been discussed in the
chapter “On the Study of Machine Learning Algorithms Towards Healthcare Appli-
cations”. Machine learning algorithms can be used to handle sophisticated and
automated multimodal biomedical data models efficiently. This chapter describes
the implementation and use of machine learning techniques in various healthcare
applications.
The chapter “A Predictive Data Analytic Approach to Get Insight of Healthcare
Databases” is aimed to discover hidden patterns from congenital heart databases
for future medical diagnosis using clustering techniques. Validation of the cited
technique has been carried out on the data based on the mortality rate for congenital
heart defects at varying age groups in India between 2006 to 2016.
The Part III of this edited volume focuses on deep learning algorithms. This section
consists of the next four chapters. Chapter “A Survey on Deep Learning Method-
ologies of Recent Applications” presents a survey on deep learning methodologies.
This chapter elaborates upon the key advancements of deep learning methodologies
in handling applications like computer vision and pattern recognition, self-driving
cars, generation of sound, creation of art, and computed predictions. A study on
privacy-Preserving recommendation system using collaborative filtering has been
discussed in the chapter “An Extensive Study of Privacy-Preserving Recommenda-
tion System Using Collaborative Filtering”. With advancement, the recommender
system is considered as one of the most information filtering-based systems that
deal with information overload by filtering vital information from large dynamically
collected information according to the user choices, interest, or item behavior. This
chapter summarizes various privacy-preserving methods used to provide security and
privacy to the user’s rating of a particular item.
A comparative study of nature-inspired algorithm based hybrid neural network
training algorithms has been discussed in the Chapter “A Comparative Study
of Nature-Inspired Algorithm Based Hybrid Neural Network Training Algorithms
in Data Classification”. The algorithm and flowchart of the whole hybridization
process of nature-inspired algorithm and standard neural network training algorithm
were elaborated, followed by a brief discussion on firefly and PSO optimization
algorithms for analysis. Chapter “Anomaly Credit Card Fraud Detection Using Deep
x Preface

Learning” introduces a deep learning model for anomaly detection for credit card
fraud in financial transactions, and further, it compares the deep learning model with
some of the existing machine learning algorithms. Validation of the proposed model
is carried out on a dataset gathered in Europe, for two days, in September 2013. The
accuracy evaluation metric is used to evaluate the proposed model’s ability to detect
credit card fraud.
The Part IV of this edited volume is on applications of deep learning techniques.
This section consists of the next three chapters. The Chapter “Application of Deep
Learning for Energy Management in Smart Grid” is based on the application of deep
learning for energy management in smart grid. The literature focuses on concepts
of load forecasting and energy management in smart grid. Different deep learning
techniques such as deep neural network, RBM, DBF, etc., based on applications
related to smart grid and smart vehicle has been elaborated.
Cost optimization of software quality assurance has been focused in the Chapter
“Cost Optimization of Software Quality Assurance”. The chapter explains that soft-
ware quality assurance constitutes an important share of an organization’s total devel-
opment overheads. The cited concepts are utilized in delivering an obliging response
on the prototypical and too funded to the frame of an acquaintance on the methods’
efficiency. Analytical approach for security of sensitive business cloud has been
discussed in the Chapter “Analytical Approach for Security of Sensitive Business
Cloud”. This chapter elaborates on the concept of security in cloud computing. It
also discusses security to sensitive data, four-level authentication authorization, data
security, network security, and cloud security. Since the increasing threat of security
in the growing demand of clouds is becoming one of the major issues, this chapter has
outlined a path to analyze and restrict dangerous security threats for cloud computing.
Numerous researchers worldwide are actively working with deep learning
concepts and integrating these concepts with computational intelligence, data anal-
ysis, cloud computing, and many other areas. The fusion of deep learning techniques
with the recent advancement area will acquire it to a newer dimension. To keep
abreast with this development in a cohesive manner, we strove to keep the book
reader-friendly. The main objective is to pen down the major developments in deep
learning areas precisely to serve as a handbook for many researchers. We believe
this edited volume will help the researchers working in deep learning areas to keep
insight into recent advances on concepts and understand the importance and usage
of deep learning techniques to handle real-life applications efficiently.

Vellore, India Debi Prasanna Acharjya


Kolkata, India Anirban Mitra
Subang Jaya, Selangor Malaysia Noor Zaman
Acknowledgment

It is with great sense of satisfaction that we present our edited book entitled “Deep
Learning in Data Analytics: Recent Techniques, Practices and Applications” and
wish to express our views to all those who helped us both direct and indirect way
to complete this project. First and foremost, we praise and heart fully thank the
almighty God, which has been unfailing source of strength, comfort, and inspiration
in the completion of this project.
While writing, contributors have referred to several books and journals, and we
take this opportunity to thank all those authors and publishers. We are extremely
thankful to the reviewers for their constant support during the process of evaluation.
Special mention should be made of the timely help given by different persons during
the project work, those whose names are not mentioned here. Last but not the least, we
thank the series editor “Kacprzyk, Janusz” and the production team of Springer-
Verlag, USA, for encouraging us and extending their cooperation and help for a
timely completion of this edited book. We trust and hope that it will be appreciated
by many readers.

xi
Contents

Theoretical Foundation of Deep Learning Theory and Analysis


A Study on Discrete Action Sequences Using Deep Emotional
Intelligence . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3
R. Santhoshkumar and M. Kalaiselvi Geetha
A Novel Noise Removal Technique Influenced by Deep
Convolutional Autoencoders on Mammograms . . . . . . . . . . . . . . . . . . . . . . . 25
Swarup Kr Ghosh, Biswajit Biswas, and Anupam Ghosh
A High Security Framework Through Human Brain Using Algo
Mixture Model Deep Learning Algorithm . . . . . . . . . . . . . . . . . . . . . . . . . . . . 45
S. Balakrishnan, J. P. Ananth, L. Ramanathan, R. Sachinkanithkar,
and S. Rajkumar
Knowledge Framework for Deep Learning: Congenital Heart
Disease . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
Ritu Chauhan and Harleen Kaur

Computing System and Machine Learning


Automatic Image Segmentation by Ranking Based SVM
in Convolutional Neural Network on Diabetic Fundus Image . . . . . . . . . . 77
Biswajit Biswas, Swarup Kr Ghosh, and Anupam Ghosh
Deep Learning in Healthcare . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 97
Deeksha Kaul, Harika Raju, and B. K. Tripathy
On the Study of Machine Learning Algorithms Towards
Healthcare Applications . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 117
Shanmuganathan Harinee and Anand Mahendran
A Predictive Data Analytic Approach to Get Insight of Healthcare
Databases . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 131
Ritu Chauhan and Harleen Kaur

xiii
xiv Contents

Deep Learning Algorithms


A Survey on Deep Learning Methodologies of Recent Applications . . . . . 145
Amit Adate and B. K. Tripathy
An Extensive Study of Privacy Preserving Recommendation
System Using Collaborative Filtering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 171
Abhaya Kumar Sahoo, Chittaranjan Pradhan, Brojo Kishore Mishra,
and Bhabani Shankar Prasad Mishra
A Comparative Study of Nature-Inspired Algorithm Based Hybrid
Neural Network Training Algorithms in Data Classification . . . . . . . . . . . 191
Sudarshan Nandy and Anirban Mitra
Anomaly Credit Card Fraud Detection Using Deep Learning . . . . . . . . . . 207
Ezz El-Din Hemdan and D. H. Manjaiah

Applications of Deep Learning Techniques


Application of Deep Learning for Energy Management in Smart
Grid . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 221
Sourasekhar Banerjee, Soumitra Ghosh, and Brojo Kishore Mishra
Cost Optimization of Software Quality Assurance . . . . . . . . . . . . . . . . . . . . 241
Soobia Saeed, N. Z. Jhanjhi, Syed Mehmood Raza Naqvi, and Azeem Khan
Analytical Approach for Security of Sensitive Business Cloud . . . . . . . . . . 257
Soobia Saeed, N. Z. Jhanjhi, Syed Mehmood Raza Naqvi, and Azeem Khan
Contributors

Amit Adate McCormick School of Engineering and Applied Science, North-


western University, Evanston, IL, United States
J. P. Ananth Sri Krishna College of Engineering and Technology, Coimbatore,
Tamilnadu, India
S. Balakrishnan Sri Krishna College of Engineering and Technology, Coimbatore,
Tamilnadu, India
Sourasekhar Banerjee Department of Computing Science, Umea University,
Umeå, Sweden
Biswajit Biswas University of Calcutta, Kolkata, West Bengal, India
Ritu Chauhan Centre for Computational Biology and Bioinformatics, Amity
University, Noida, Uttar Pradesh, India
M. Kalaiselvi Geetha Department of Computer Science and Engineering, Anna-
malai University, Tamilnadu, India
Anupam Ghosh Netaji Subhash Engineering College, Kolkata, West Bengal, India
Soumitra Ghosh Department of Computing Science and Engineering, Indian
Institute of Technology, Patna, India
Swarup Kr Ghosh Sister Nivedita University, Kolkata, West Bengal, India
Shanmuganathan Harinee Vellore Institute of Technology, Vellore, India
Ezz El-Din Hemdan Department of Computer Science and Engineering, Faculty
of Electronic Engineering, Menoufia University, Menouf, Menoufia, Egypt
N. Z. Jhanjhi School of Computing and IT, Taylor’s University, Subang Jaya,
Malaysia
Deeksha Kaul SCOPE, VIT, Vellore, Tamil Nadu, India
Harleen Kaur Department of Computer Science and Engineering, School of
Engineering Sciences and Technology, Jamia Hamdard, New Delhi, India
xv
xvi Contributors

Azeem Khan ADP, Taylor’s University, Subang Jaya, Malaysia


Anand Mahendran Vellore Institute of Technology, Vellore, India
D. H. Manjaiah Department of Computer Science, Mangalore University, Manga-
lore, India
Bhabani Shankar Prasad Mishra School of Computer Engineering, Kalinga
Institute of Industrial Technology, Bhubaneswar, India
Brojo Kishore Mishra Department of Computer Science and Engineering, GIET
University, Gunupur, India
Anirban Mitra Amity School of Engineering and Technology, Amity University,
Kolkata, India
Sudarshan Nandy Amity School of Engineering and Technology, Amity Univer-
sity, Kolkata, India
Syed Mehmood Raza Naqvi Department of Computer Science, Sheridan College,
Brampton, Canada
Chittaranjan Pradhan School of Computer Engineering, Kalinga Institute of
Industrial Technology, Bhubaneswar, India
S. Rajkumar School of Computer Science and Engineering, VIT, Vellore, India
Harika Raju SCOPE, VIT, Vellore, Tamil Nadu, India
L. Ramanathan School of Computer Science and Engineering, VIT, Vellore, India
R. Sachinkanithkar Sri Krishna College of Engineering and Technology, Coim-
batore, Tamilnadu, India
Soobia Saeed Department of Software Engineering, Universiti Teknologi, Johor
Bahru, Malaysia
Abhaya Kumar Sahoo School of Computer Engineering, Kalinga Institute of
Industrial Technology, Bhubaneswar, India
R. Santhoshkumar Associate Professor, Department of Computer Science and
Engineering, St. Martin’s Engineering College, Secunderabad, Telangana, India
B. K. Tripathy SITE, VIT University, Vellore, Tamil Nadu, India
Acronyms

ABF Adaptive Bilateral Filter


AD Alzheimer’s Disease
AE Autoencoder
AI Artificial Intelligence
ALE Arcade Learning Environment
AMR ANN Misclusting Rate
ANN Artificial Neural Network
API Application Program Interface
BBIV Block-Based Intensity Value
BICC Block Intensity Comparison Code
BM Boltzmann Machine
BN Batch Normalization
BP Belief Propagation
CAD Computer Aided Detection
CAE Convolutional Autoencoder
CANFIS Coactive Neuro Fuzz Inference System
CART Classification and Regression Tree
CDA Convolutional Denoising Autoencoder
CDAEM Conventional Denoising Autoencoder Method
CDC Centers for Disease Control
CF Collaborative Filtering
CHAID Chi-square Automatic Interaction Detection
CHD Coronary Heart Disease
CI Confidence Interval
CNN Convolutional Neural Network
COHD Congenital Heart Defect
CV Cross Validation
CVD Cardio Vascular Disease
DAE Denoising Autoencoder
DAEM Denoise Autoencoder Method
DBL Different Bin Level
DBM Deep Boltzmann Machine
xvii
xviii Acronyms

DBN Deep Belief Network


DCNN Deep Convolutional Neural Network
DES Data Encryption Standard
DGM Deep Generative Model
DL Deep Learning
DM Data Mining
DNN Deep Neural Network
DP Dynamic Programming
DR Dimensionality Reduction
DRIVE Digital Retinal Images for Vessel Extraction
DSC Dice Similarity Coefficient
DSM Demand Side Management
DSOR Decision Support and Operations Research
DT Decision Tree
DWT Discrete Wavelet Transform
EEG Electroencephalography
EV Electric Vehicle
FDCNN Feature Difference Convolutional Neural Network
FEL Feature and Ensemble Learning
FFBPNN Firefly based Back Propagation Neural Network
FMRI Functional Magnetic Resonance Imaging
FNN Feedforward Neural Network
GA Genetic Algorithm
GABNN Genetic Algorithm based Neural Network
GAN Generative Adversarial Network
GEMEP Geneva Multimodal Emotion Portrayals
GO Gene Ontology
GRNN General Regression Neural Network
GRU Gated Recurrent Unit
HCI Human–Computer Interaction
HE Homomorphic Encryption
HER Human Emotion Recognition
HFEN High Frequency Error Norm
HOG Histograms of Oriented Gradient
HRBVS Hierarchical Retinal Blood Vessel Segmentation
IBEF Indian Brand Equity Foundation
IDE Integrated Development Environment
IT Information Technology
ITBN Interval Temporal Bayesian Network
JLT Johnson-Lindenstrauss Transformation
JSC Jaccards Similarity Coefficient
KDD Knowledge Discovery of Database
KLD Kullback-Leibler Divergence
KNN k-Nearest Neighbor
KPCA Kernel Principal Component Analysis
Acronyms xix

LDA Linear Discriminant Analysis


LDL Low Density Lipoprotein
LIDAR Light Detection and Ranging
LIF Linear Index of Fuzziness
LLC Locality constrained Linear Coding
LMS Local Motion Signatures
LR Logistic Regression
LSTM Long Short Term Memory
LTLF Long-Term Load Forecasting
MAE Mean Absolute Error
MAPE Mean Absolute Percentage Error
MCI Mild Cognitive Impairment
MD Medical Device
MDP Markov Decision Process
MEBN Multi Entity Bayesian Network
MI Machine Intelligence
ML Machine Learning
MLP Multilayer Perceptron
MMDML Modified Multi-manifold Deep Metric Learning
MPP Motion Projection Profile
MRI Magnetic Resonance Imaging
MRP Modified Random Perturbation
MSE Mean Square Error
MTLF Mid Term Load Forecasting
MTTF Mean Time to Failure
NAS Neural Artistic Style
NC Normal Control
NI Nature Inspired
NIN Network in Network
NLM Non Local Mean
NMF Non-negative Matrix Factorization
NPDR Non Proliferate Diabetic Retinopathy
NRI Nonreceived Issue
PAC Probably Approximately Correct
PC Personal Computer
PCA Principal Component Analysis
PD Parkinsons Disease
PDR Proliferate Diabetic Retinopathy
PET Positron Emission Tomography
PGNN Parallel Granular Neural Network
PIN Personal Identification Number
PNN Probabilistic Neural Network
PPGN Plug and Play Generative Network
PPNBC Privacy-Preserving Naive Bayesian Classification
PPV Positive Predictive Value
xx Acronyms

PQS Picture Quality Score


PSAW Predictive Situation Awareness
PSNR Pick Signal-to-Noise Ratio
PSO Particle Swarm Optimization
QA Quality Assurance
RBF Radial Basis Function
RBM Restricted Boltzmann Machine
RBVS Retinal Blood Vessel Segmentation
ReLU Rectified Linear Unit
RF Random Forest
RGB Red Blue Green
RMSE Root Mean Square Error
RNN Recurrent Neural Network
ROC Receiver Operating Characteristic
ROI Region Of Interest
RPT Randomized Perturbation Technique
RRT Randomized Response Technique
RSVM Rank Support Vector Machine
SBP Systolic Blood Pressure
SDAE Stacked Denoising Auto Encoders
SDBP Steepest Descent Back Propagation
SIFT Scale-Invariant Feature Transform
SMOTE Synthetic Minority Oversampling Technique
SOM Self Organizing Map
SQA Software Quality Assurance
SSA Software Superior Affirmation
SSIM Structural Similarity Index Metric
SSL Secure Sockets Layer
STLF Short-Term Load Forecasting
SVD Singular Value Decomposition
SVM Support Vector Machine
SVR Support Vector Regression
TLS Transport Layer Security
UM Uniformity Measure
UQI Universal Quality Index
VFD Variable Frequency Drive
VMM Virtual Machine Monitor
Theoretical Foundation of Deep Learning
Theory and Analysis
A Study on Discrete Action Sequences
Using Deep Emotional Intelligence

R. Santhoshkumar and M. Kalaiselvi Geetha

Abstract Recognition of emotions from human plays a vital role in our day to day
life and is essential for social communication. Automatic emotion recognition is
becoming recent research focus on artificial intelligence. A facet of human intel-
ligence is the ability to recognize emotion that is regarded as one of the attribute
of emotional intelligence. Although research based on facial expressions or speech
is seen in thrive, recognizing emotions from body gestures remains a less explored
topic. This chapter proposes a machine learning approach and discussed with deep
learning model to achieve emotional intelligence. The block based intensity value
(BBIV) feature and the different bin level HoG feature (DBLHoG) are extracted
from human body movements and are fed to a supervised learning algorithm. Sup-
port vector machine (SVM), k-nearest neighbor (KNN) and random forest classifiers
are the supervised learning algorithm used in this chapter. Finally, the pre-trained
deep convolutional neural network (DCNN) model is used. The experiment is con-
ducted using Geneva multimodal emotion portrayals (GEMEP) corpus dataset. In
this dataset, human body movement expressing the five archetypical emotions likes
(anger, fear, joy, pride and sad). In this emotions recognition system, The random
forest classifier outperformed better than the SVM and kNN classifier. Finally DCNN
model achieve better recognition than random forest classifier. This chapter gives a
brief study on achieving emotional intelligence with a DCNN Model.

1 Introduction

The ability to understand the human emotions plays a central role human social
behaviour and interaction. The emotions are important in decision making and ratio-

R. Santhoshkumar (B)
Associate Professor, Department of Computer Science and Engineering, St. Martin’s Engineering
College, Secunderabad 500100, Telangana, India
e-mail: [email protected]
M. K. Geetha
Department of Computer Science and Engineering, Annamalai University, Tamilnadu, India
e-mail: [email protected]

© Springer Nature Switzerland AG 2022 3


D. P. Acharjya et al. (eds.), Deep Learning in Data Analytics, Studies in Big Data 91,
https://doi.org/10.1007/978-3-030-75855-4_1
4 R. Santhoshkumar and M. K. Geetha

nal thinking on experimental psychology in recent research area. Over the years
research in emotion recognition mainly concentrated on facial expression, voice
analysis, full body movements and gestures. In different way human beings express
different type emotions for day to day communications. Understanding human emo-
tions is a key area of research, since recognizing emotions may provide a plethora of
opportunities and applications for instance, friendlier human-computer interactions
with an enhanced communication among humans, by refining the emotional intelli-
gence. The human communication includes verbal and non verbal communication.
The non-verbal communication is a sharing of wordless clues or information using
eyes, hands, heads, fingers and all body movements. This includes visual cues such
as body language and physical appearance. Human Emotion can be recognized using
body language and posture. Posture gives information which is not present in speech
and facial expression. For example, the emotional state of a person from a long
distance can be identified using human posture. Hence human emotion recognition
through non-verbal communication can be achieved by capturing body movement.
This chapter aims to examine the motion cues that indicate difference between
emotions. The possibility to use motion and gestures as indicators of the state of indi-
viduals provides a novel approach to quantitatively monitor and estimate the users in
an ecological environment and to react adaptively to them. This approach recognizes
emotional states such as anger, joy, sad, fear etc. based on body movement analysis.
From the recent literature it is evident that different emotions are often associated with
various body movements. The speed, amplitude and fluidity of movement express
specific emotions from human body movement. The analysis of emotional behavior
that provides indicators describing the dynamics of expressive motion cues using
static actions. The communication greatly improve by understanding and knowing
how to respond to peoples expression. The human communication includes not only
the language spoken, but also non-verbal cues as movements of hands, head and
body gestures. The important role in psychological research area is to develop the
concepts that may support the HCI (Human Computer Interaction) technologies and
understanding human emotions [1, 2]. The fundamental emotions like anger, neutral,
happy, fear, disgust, sadness, surprise, etc., are the particular emotions are related to
certain human body movements. For example: joy brings upward acceleration of the
fore arms and make body to openness, fear gives to contrast the body, the signal of
fear and sadness brings body turning away. Body turning towards indicates happi-
ness, anger, surprise. A recent survey reviews the literature on emotion recognition
from body posture and movement, and discusses the main challenges in collecting
appropriate data sets and ground truth labelling. Both automatic recognition of emo-
tion and generation of affect expressive movements are developed by computational
model. The emotion recognition systems provides applications in several motivat-
ing areas: in surveillance field, estimate emotional state of students in intelligent
tutoring systems, social robotics for social interaction, monitor player’s motivation
and interest in games, in medical field to monitoring depression levels of patients
and provide applications for autism and dementia patients. Some other application
domains are, educational software, telecommunications, automobile safety, video
games, animations, psychiatry, robotics, affect sensitive HCI [1, 3].
A Study on Discrete Action Sequences Using Deep Emotional Intelligence 5

Ashwini et al. describes the different types of approaches and developed a real time
emotion recognition system used for recognizing human emotions [1]. Stefano et al.
proposed automatic emotion recognition in real time from body movements. The real
time video are captured and converted into 3D skeletal frames using advance d video
capturing system. From the sequences of 3D skeletons, the kinematic, geometrical
and postural features are extracted and given to the multi class SVM classifier to
categorise the human emotion [3]. A survey on generation of such body movements
and the state of the art on automatic recognition of emotion is also presented in the
literature [2]. The important characteristics such as the representation of affective
state, the body movements analyzed and the use of information systems are dis-
cussed. A framework for behaviour recognition from human upper body movements
is presented in the literature [4]. The reduced amounts of visual information are used
to analysis the affective behaviour of body movements. The aim of the work is to
individuate a representation of emotional displays depending on nonverbal gesture
features. Further an advanced real time system for human body movements to rec-
ognize emotions continuously is presented [5]. The high-level kinematic features,
geometrical features and the united 3D postural features are given as input to ran-
dom forests classifier. Furthermore, a system for recognition of emotion depends
on different actions, different expression of emotions and low level body cues from
human body movement. To recognition the emotion from these aspects, the features
are extracted from the various parts and are fed to the random forest classifier [6].
Besides, a body posture and body action coding system from body movement on
an anatomical level of different articulations of body parts, direction and orientation
of movement is developed [7]. Similarly, a survey on recognizing human emotion
from hand, arms, gestures and body movements is conducted [8]. Further a robust
technique for assessing human body expression based on movement characteristics
with positive and negative emotions is developed [9]. In the literature, the study and
analysis of the spatial and temporal information structure of the motion capture data
and extract features that are related to affective state descriptors is also discussed [10].
Again a survey on recent advances in developing robust techniques and modalities
for automatic human emotion recognition system from body movements is also
presented [11]. Here the importances of body movement segmentation are discussed
and advanced application areas are described. Likewise, an analysis of emotional
behaviour system based on classification of time series and dynamics of expressive
motion cues is also proposed [12]. Similarly, a video retrieval applications is proposed
for shot detection and video classification using block intensity comparison code
(BICC) and unsupervised shot detection [13]. It uses a noval ANN misclusting rate
(AMR) algorithm to detect the shot transitions. Similarly, an automatic depression
analysis system from human gestures and upper body expressions is also developed
[14]. The bag of words and space time interest points is developed for the analysis
of facial and upper body movements.
Furthermore a robust gesture recognition system using learning local motion sig-
natures (LMS) is proposed [15]. Similarly, the human detection and robust visual
object recognition using adopting linear SVM is discussed in the literature [16]. It
showed the performance of human detection using feature sets of histograms of ori-
6 R. Santhoshkumar and M. K. Geetha

ented gradient (HOG) descriptors. Further, a quantized trajectory snippets method


for tracked features is proposed [17]. It is a simple feature tracking method and com-
putationally efficient for motion detection. Arunnehru and Kalaiselvi [18] developed
motion projection profile (MPP) feature on region of interest (ROI) for automatic
activity recognition from body movement. In addition a real time vision systems
for video analysis in describing scale-invariant feature transform (SIFT) feature
extraction algorithms and novel implementations of the KLT feature tracking is
proposed [19]. To move further, a motion intensity for real time action recognition in
surveillance video using ROI from the difference image is developed [20]. Besides,
an automatic human emotion recognition in surveillance video based on gesture
dynamic’s features is developed and the features are evaluated by SVM, Nave Bayes
and dynamic time wrapping [21].
In this chapter, an emotion recognition system is proposed based on BBIV feature
and HOG feature with different bin levels. Finally, the supervised learning model is
compared with DCNN model. First, the video are converted into RGB, and then it
converted into grey frame. Second, plot the bounding box for human present in the
frame using (x, y) co-ordinates and the center of the frame. For BBIV feature extrac-
tion, subtract the two consecutive frames. The difference frame is further divided into
three blocks and the maximum motion block is further divided into (6 × 6) blocks.
Using different bin levels extract, the HOG feature for the odd number frames. Then,
the feature points are stored as training sample and extract same feature points for
testing set and stored it as testing sample. Finally, recognize the emotions using ran-
dom forest, SVM and kNN classifier. The experiments are conducted using GEMEP
database.
The rest of the chapter is organized as follows: Following introduction, emotion
recognition from static action sequences is presented in Sect. 2. It is further followed
by feature descriptions in Sect. 3. Section 4 presents preformance evaluation followed
by deep emotional intelligence in Sect. 5. the chapter is concluded in Sect. 6.

2 Emotion Recognition from Static Action Sequences

Emotion recognition approaches are normally based on motion identification from


video sequence. The following discussions briefly explain the approaches for motion
detection and image feature representation. This is helpful to discriminate human
emotion from various levels of granularity. Emotion refers to the intricate sequence
of motion patterns performed in a restricted style. For example, the emotion for the
phrase ‘angry’ indicates the sequences of motion patterns such as forward head,
forward arms and forehand. Similarly, recognizing a human emotion really involves
classifying basic emotions such as angry, fear, joy, pride and sad. Thus the emotion
recognition approaches naturally follows a hierarchical pattern. The following Fig. 1
depicts an overview of proposed research on emotion recognition.
In the proposed model, the input videos are converted into RGB frames and it is
converted into gray frames. To detect the human in the frame, the gray frames are
A Study on Discrete Action Sequences Using Deep Emotional Intelligence 7

Fig. 1 An overview of the proposed approach on emotion recognition

Fig. 2 Bounding box of human detection in one frame

converted into binary frames and then put bounding box around human body. These
processes are called as pre-processing. From the sequences of gray frame the different
bin level HoG (DBLHoG) features and bock based intensity value (BBIV) features
are extracted and stored separately. The k-nearest neighbor and random forest are
employed to classify the trained and test video to recognize the human emotion. The
pre-processing block converts videos into RGB frames and RGB frames into gray
frames of all input videos. The human detection from gray frame using bounding
box is shown in Fig. 2 for angry emotion.

3 Feature Descriptions

This section describes the features associated with human emotion. The human basic
emotions such as angry, joy, fear, sad and pride are consider for this experiment from
GEMEP corpus dataset. The overall structure of the proposed work is shown in Fig. 3.
8 R. Santhoshkumar and M. K. Geetha

Fig. 3 Overview of the proposed approach

3.1 Block Based Intensity Value Feature Extraction

Initially the input video are preprocessed and converted into frames in BBIV fea-
ture extraction. Frame subtraction is a subtraction of successive frames in video
sequences. It is used for change detection. The difference image is produced by two
input frames at time t and (t + 1). The high amplitude regions shown in difference
image are considered as motion regions and is defined in Eq. 1.

Dk = |I ntk (i, j) − I ntk+1 (i, j)|; 1 ≤ i ≤ w; 1 ≤ j ≤ h (1)

In Eq. 1, Dk (i, j) refers to the difference image, I ntk (i, j) is the intensity of the
pixel (i, j) in the kth frame, w and h are the width and height of the image respectively.
The motion region is considered as the region of interest (ROI). Figure 4 shows the
two successive frames of the GEMEP dataset and shows the frame differencing
image. Motion information, M I n f ok is calculated using Eq. 2.

1 if Dk (i, j) > t
M I n f ok (i, j) = (2)
0 Otherwise

The input videos are preprocessed and converted into n frames. The motion infor-
mation is calculated from difference image and is considered as ROI. The difference
image of size (720 × 360) is divided into three blocks B1, B2, and B3 and each of
pixel size (240 × 360). Then the block having maximum intensity value is further
divided into (6 × 6) block each of pixel size (40 × 60). The 36 dimensional fea-
ture vectors are extracted from the (6 × 6) block. The Fig. 5 demonstrates the block
division and 36 dimensional feature extraction blocks.

3.2 Different Bin Level HoG Feature Extraction

The HoG feature is defined as local object appearance and shape can often be charac-
terized rather well by the distribution of local intensity gradients of the corresponding
A Study on Discrete Action Sequences Using Deep Emotional Intelligence 9

Fig. 4 A difference frame

Fig. 5 Block division

gradient [16, 22]. It is said that the description of the HOG method that has been used
in its higher form is scale invariant features transformation (SIFT) and it has been
broadly demoralized in human detection. The proposed architecture of DBLHOG
feature extraction is depicted in Fig. 6.
The subsequent building of a 1D histogram whose concatenation supplies the
feature vector from the HOG descriptor using gradient directions among the pixels
in the cell. The image to be analyzed as intensity function L. The image is further
divided into cells of size (3 × 3) pixels with different histogram bins (9, 15, 20, 25,
30). The Eqs. 3 and 4 define the gradient magnitude g and the gradient orientation θ .
10 R. Santhoshkumar and M. K. Geetha

Fig. 6 Proposed architecture of DBLHOG feature

These equations are used to compute magnitude g and orientation θ for all the pixels
in the block from the image gradients.

g(a, b) = gx (a, b)2 + g y (a, b)2 (3)

g y (a, b
Θ(a, b) = arctan (4)
gx (a, b)

Further, compute a feature vector Vi j for each cell ci j in the block. The Eq. 5
defines the weighted gradient magnitude by quantizing the unsigned orientation into
K orientation bins.
Vi j = [Vi j (β)]T (5)

Vi j = g(a, b)δ[bin(a, b) − β] (6)
(a,b)∈ci j

The index of the orientation bin with the pixel (a, b) returns the function bin(a, b)
and the function δ[] is the Kronecker delta. The coefficient ρ normalize the feature
vector in all cells from 2D descriptor of block. The extraction of HoG feature from
gray image is depicted in Fig. 7.


3 
3 
K
ρ= Vi j β (7)
i=1 j=1 β=1

In order to obtain a n-dimensional feature vector, the different bin level feature
values are concatenated into a distinctive vector. Using this approach the features
are extracted for different bin level such as (9, 15, 20, 25, 30) of training and testing
dataset. The extracted features are modeled by the KNN and random forest classifiers
for emotion detection.
A Study on Discrete Action Sequences Using Deep Emotional Intelligence 11

Fig. 7 Extraction of HoG feature from gray image

3.3 K Nearest Neighbor

K -nearest neighbor is a type of supervised learning method. It is a non-parameter


algorithm where samples are classified depending on the category of their near-
est neighbor. According to it, the classification algorithm finds the test sample’s
categories which are the nearest neighbor to the test sample. However, the classifi-
cation algorithm needs to compute all distance between training and testing sample.
The process of KNN algorithm categorize sample S. Assume q training samples
A1, A2, . . . , Aq. After feature reduction N is the addition of the training samples
and get n-dimension feature vector. The all training samples (S1, S2, . . . , Sn) have
the same feature vector of sample S and evaluate the similarities among them. For
example taking the pth sample b p (b p1 , b p2 , . . . , b pn ), the similarity is defined as
S I M(S, b p ). n
q=1 Sq , b pq
S I M(S, b p ) =   n (8)
n 2
( q=1 Sq )2 q=1 b pq


R(S, Aq ) = b(S, b p )y(b p , Aq )

The larger N similarities, S I M(S, b p ), p = 1, 2, . . . , N , of k samples are chosen


and consider them as a K nearest neighbor collection of S. Then, the probability of
S is calculated.
12 R. Santhoshkumar and M. K. Geetha

3.4 Random Forest Classifier

The random forest (RF) is the another type of supervised machine learning algorithm.
There is a relationship between results obtained and number of trees in the forest.
The two steps in this algorithm are [23] random forest creation and random forest
prediction. The procedure for RF creation and prediction is given below.
Algorithm 1 Random Forest Creation
1. Select “k” features randomly from total “m” features, where k < m
2. Using best split point calculate the node “d” from the “k” features
3. Using best split, divide node into daughter node
4. Repeat the steps 1 to 3, until one number of nodes has been reached
5. To create n number of trees, repeating the steps 1 to 4
Algorithm 2 Random Forest Prediction
1. Calculate the votes for each prediction
2. Final prediction from this algorithm can be considered by high voted prediction

3.5 Support Vector Machine

The support vector machine (SVM) is an important and efficient technique for clas-
sification in visual pattern recognition [20]. The SVM is most extensively used in
kernel learning algorithm. The elegant theory used to separate two classes by large-
margin hyperplanes. It cannot be extended easily to separate N mutually exclu-
sive classes. The most popular “one-vs-others” approach is used for the multi class
problem where, one class is separated from N classes. The classification task typ-
ically involves with training and testing data. The training data are separated by
(s1, t1), (s2, t2), . . . (sn, tn) into two classes, where bj ∈ 1, −1 are the class labels
and s j ∈ tn contains n-dimensional feature vector. The goal of SVM is to develop a
model which predicts target value from testing set. w.s + b = 0 is the hyper plane of
binary classification, where w ∈ R N . M = 2/||w|| is the large margin. The Lagrange
multipliers αi, i = 1, . . . , m are used to solve the minimization problem, where v
and y are optimal values obtained from Eq. 9.


n
h(s) = sgn x j b j L(S j , s) + y (9)
j=1

1  k
min v,y, v R v + D j (10)
2 j=1

b j V R φs j + y ≥ 0 (11)
A Study on Discrete Action Sequences Using Deep Emotional Intelligence 13

The Eqs. 10 and 11 obtain the soft margin classifier. When the training sample
is not linearly separable, the input space mapped into high dimensional space using
kernel function. The multiclass SVM is constructed by N -binary classifiers and one
class was separated from rest of the class. Here “one-vs-others” approach is used
in this SVM. The five classes of emotions are used in this work. The jth class of
the training sets have positive labels and all others with negative labels. Finally, the
feature vectors from the body movement feature are given into multiclass SVM for
classification of human emotion.

3.6 GEMEP Dataset

The GEMEP is a set of audio and video recordings. The 18 affective states of emo-
tional expression can be acted by 10 actors. They acted in various types of expression
and verbal contents. From that five basic emotions, such as angry, joy, fear, sad and

Fig. 8 Example frames of five basic emotions (Angry, Fear, Joy, Pride and Sad)
14 R. Santhoshkumar and M. K. Geetha

pride have been chosen for this work. There are 10 actors, 5 male and 5 female were
acted in each emotion videos. The resolutions of the recorded videos are (720 × 576)
and each video has 25 frames per second (fps) [24]. The data set is depicted in Fig. 8.
This experiment is conducted in MATLAB R2015a in computer with windows 7
operating system and Intel Xeon X3430 processor 2.40 GHz with 8 GB of RAM. The
feature extraction techniques are explained in above section, using n-dimensional
DBLHoG features and BBIV features are extracted. The metrics of the proposed
feature is tested on SVM, KNN and RF classifiers.

4 Performance Evaluations

According to the histogram bins, different sets of features were extracted with dif-
ferent dimensional levels. There are five different dimensional levels of feature were
used. Those extracted features are given to the SVM, KNN and RF classifiers one by
one. Accuracy, recall, F-score, specificity and precision are the measuring assessment
for this execution. The factual evaluation of accuracy, recall, F-score, specificity and
precision are given as below.
The various measures of confusion matrix is defined with true positives (t p), true
negatives (tn), false positives ( f p), and false negatives ( f n). The ratio between sum
of correct classifications and total number of classifications is called as accuracy. It
is given as:
t p + tn
Accuracy =
tn + f p + t p + f n

The ratio between correctly labeled instances and total instances in the class is
said to be recall. Similarly, the ratio between correctly labeled instances and total
labeled instances is called as precision. It is a percentage of positive predictions in
specific class that are correct. A good classifier can provide both recall and precision
values high. The harmonic mean of precision and recall is called as F-measure. These
measures are defined below.
tp
Pr ecision =
tp + f p

tp
Recall =
tp + f n

Pr ecision × Recall
F − Scor e = 2
Pr ecision + Recall
A Study on Discrete Action Sequences Using Deep Emotional Intelligence 15

4.1 Performance Measure of BBIV Feature with KNN, RF


and SVM

The confusion matrix of KNN, RF and SVM classifier are shown in Tables 1, 2, and 3
respectively. The percentage of instance that are classified accurately are illustrated
in the diagonal of confusion matrix. The each emotion class occurrence is spoken
to by the lines and the emotion class anticipated by the classifier is spoken to by the
sections. The emotions like sad, fear and pride are grouped well with precision more
noteworthy than 90% in kNN, 92% in RF and 88% in SVM. From this, angry and
joy emotions are confused as curve, where these two emotions instinctively appear
to be difficult to recognize.

4.2 Performance Measure of DBLHOG Feature with KNN,


RF and SVM

The confusion matrix of KNN, RF and SVM classifier with respect to DBLHOG is
shown in Tables 4, 5, and 6 respectively. The percentage of instance that are classified
accurately illustrated in the diagonal of confusion matrix. The each emotion class
occurrence is spoken to by the lines and the emotion class anticipated by the classifier
is spoken to by the sections. The emotions like sad, fear and pride are grouped well
with precision more noteworthy than 80% in KNN, 93% in RF, and 85% in SVM.

Table 1 Confusion matrix for BBIV feature in kNN Classifier


Emotion Angry Fear Joy Pride Sad
Angry 90.4 3.1 3.7 2.5 0.3
Fear 1.7 90.8 3.9 2.6 1.0
Joy 1.5 1.4 95.7 1.0 0.4
Pride 1.2 3.7 1.9 92.8 0.4
Sad 0.0 1.8 0.0 0.0 98.2

Table 2 Confusion matrix for BBIV feature in RF classifier


Emotion Angry Fear Joy Pride Sad
Angry 92.4 3.1 2.6 1.6 0.3
Fear 0.9 92.8 3.9 2.4 0.0
Joy 0.5 1.6 96.9 1.0 0.0
Pride 0.2 3.6 1.9 93.9 0.4
Sad 0.0 0.0 0.0 0.0 100
16 R. Santhoshkumar and M. K. Geetha

Table 3 Confusion matrix for BBIV feature in SVM classifier


Emotion Angry Fear Joy Pride Sad
Angry 91.8 4.0 1.6 0.5 2.2
Fear 0.9 93.2 1.1 0.9 3.9
Joy 0.9 5.3 89.8 0.8 2.1
Pride 1.7 4.2 4.4 82.7 7.0
Sad 0.0 0.3 0.0 0.0 99.7

Table 4 Confusion matrix for DBLHOG feature in KNN Classifier


Emotion Angry Fear Joy Pride Sad
Angry 83.1 4.5 8.4 3.3 0.8
Fear 8.1 78.2 8.5 4.4 0.9
Joy 9.5 7.4 77.7 4.2 1.1
Pride 12.9 9.1 7.8 68.6 1.7
Sad 3.3 1.8 2.3 1.3 91.2

Table 5 Confusion matrix for DBLHOG feature in RF classifier


Emotion Angry Fear Joy Pride Sad
Angry 93.2 4.3 0.6 3.3 1.9
Fear 0.4 93.9 0.0 4.4 5.7
Joy 1.0 4.2 92.6 4.2 1.9
Pride 0.0 3.0 2.1 68.6 6.5
Sad 0.0 0.3 2.3 0.0 99.7

Table 6 Confusion matrix for DBLHOG feature in SVM classifier


Emotion Angry Fear Joy Pride Sad
Angry 80.2 6.5 7.3 4.6 1.4
Fear 1.6 88.6 4.3 2.6 2.9
Joy 3.3 3.6 85.5 3.8 3.7
Pride 3.1 3.7 2.7 87.1 3.4
Sad 2.7 1.8 2.5 2.7 90.3

From this, angry and joy emotions are confused as curve, where these two emotions
instinctively appear to be difficult to recognize.
A Study on Discrete Action Sequences Using Deep Emotional Intelligence 17

5 Deep Emotional Intelligence

Humans possess the ability to address problems, gain knowledge, create ideas, rec-
ognize patterns, provide decisions, reason, plan and think through their intelligence.
Intelligence is the intellectual skill of humans, which is discernible by high levels
of enthusiasm and self awareness. Artificial intelligence is the building of human
abilities artificially. The artificial intelligent system explores to some extent creative
thinking; learning and decision making of humans. The current focus of research
is permitting machines to shape the globe effectively to demonstrate intelligence
system. A large amount of data has to be stocked, explicitly or implicitly to realize
this system. Since, researchers are trying to create learning algorithms to capture
knowledge from that information. A lot of challenges of artificial intelligence are
considered for developing the learning algorithms.
The most important aspect for intelligent machines is considered as learning. Arti-
ficial intelligence is executed by machine learning technique. In artificial intelligence
research, a machine teaches to detect various patterns using machine learning pattern.
Conventional machine learning techniques expertise to design a feature extractor that
transformed the raw data into a feature vector could detect or classify patterns in the
input data. Deep learning is a dedicated form of machine learning. The technique
that instructs computers to do some operation and behave like humans is done by
machine learning. In machine learning methods the feature extraction methods starts
from input data. The features are fed to model that classifies the objects in the image.
Learning feature hierarchies are produced by combining of lower level and higher
level features [25]. In deep learning model, features are automatically extracted from
inputs data by several levels of abstraction. An overview of learning based methods
is depicted in Fig. 9.

Fig. 9 Learning based methods


18 R. Santhoshkumar and M. K. Geetha

The various types of machine learning methods are ever rising at an enormous
rate. Deep learning has emerged as a popular approach within machine learning.
The two major methods in human emotion recognition (HER) problem are machine
learning and deep learning methods [26]. The several limitations of machine learning
methods are failure to learn features automatically from the input data, and deep rep-
resentation of data in classifiers. Alternatively, a deep learning-based approach plays
the concept of end-to-end learning by using the trainable feature extractor followed
by a trainable classifier. The multiple layers of features are automatically extracted
from the raw data. This deep learning algorithm develops multi layer representation
of different patterns in the input data, where each successive layer is responsible for
learning increasingly complex features [27]. The lower layers extracting higher level
features from the input data, thus representation increased at abstraction level and
at each consecutive layer. The automatic learning ability of deep learning method
neglected the need of handcrafted feature detectors and descriptors. In many visual
categorization tasks, the deep learning models have shown higher performance than
traditional handcrafted feature-based techniques. The convolutional neural networks
(CNN) and deep belief networks (DBN), deep recurrent neural networks (RNN), and
deep Boltzmann machines (DBM) are the deep learning model employed for many
visual categorization tasks. DBN is an unsupervised probabilistic graphical model
capable of learning from the input data without any prior knowledge. This model can
also be trained in a semi-supervised or unsupervised fashion which is quite helpful
of labeled data or dealing with unlabeled data. The non-deep learning approach and
deep learning approach are the two learning based methods [28].

5.1 Non-Deep Learning Based Methods

Non deep learning based methods are dictionary learning method and genetic pro-
gramming method. Briefly, these methods are discussed below.
Dictionary Learning Method The emotion recognition, action recognition,
image and video classification are the computer vision applications, which can
be experimented using dictionary learning based approaches [29]. From the large
number of samples, the representative vectors are learned and used in this concept.
Further, a framework is developed for human action recognition using dictionary
learning methods [30]. Based on the hierarchical descriptor the proposed method
[31] for human activity recognition outperforms the state-of-the-art methods. For a
visual recognition, a cross domain dictionary learning based method was developed.
An unsupervised model further developed for cross view human action recognition
[32] without any label information. The coding descriptors of locality constrained
linear coding (LLC) [33] are generated by a set of low level trajectory features for
each action.
Genetic Programming Approach The unknown primitive operations can
improve the accuracy performance of the human emotion recognition task by using
genetic programming technique. Now a day, this type of approach was introduced
A Study on Discrete Action Sequences Using Deep Emotional Intelligence 19

for emotion recognition [34]. In this approach, the spatio-temporal motion features
are automatically learned for action recognition. The 3D Gabor filter and wavelet are
evolved for this motion feature. Similarly, the valuable set of features was learned
for emotion recognition.
Probabilistic Graphical Model A probabilistic graphical model indicates the
dependencies and random variables in a directed acyclic graph form. The conditional
Bayesian networks, temporal Bayesian networks, and multi entity Bayesian networks
(MEBN) are the different types of Bayesian network. An interval temporal Bayesian
networks (ITBN) was introduced for recognition of complex human activities [35].
In order to evaluate the performance of the proposed method, a cargo loading dataset
was considered for experimentations and evaluations. Similarly, another method
was proposed for action detection using dynamic conditional Bayesian network,
which also achieved the state-of-the-art results. In MEBN, the predictive situation
awareness (PSAW) using multiple sensors is used [36]. For predicting and estimating
the temporally evolving situations these types of networks are robust for reasoning
the uncertainty in the complex domains.

5.2 Deep Learning Based Model

There are no most excellent hand crafted feature descriptors for all types of dataset.
To handle these types of problem the features are directly learning from raw data.
Learning multiple levels of representation in data such as speech, images, videos, and
text is more advantageous in deep learning. These models have automated feature
extraction, classification and process the images as raw data. These models have
multiple processing layers in this work. There are three types of approaches in deep
learning models [37]. These are generative or unsupervised approach, discriminative
approach, and hybrid models (the characteristic combination of both approachs).
Generative or Unsupervised Approach The class labels are not required for
learning process in the unsupervised deep learning method. These types of methods
are specifically useful when labeled data are relatively unavailable. A remarkable
surge in the history of deep models was triggered by the work of Hinton [38]. In an
unsupervised pre-training learning stage, a back propagation method is used for fine
tuning. These types of deep learning approaches are used for many applications like
object identification, image classification, speech classification, activity and emotion
recognition. An unsupervised feature learning models from video data was proposed
for human action recognition [39]. The authors used an independent subspace anal-
ysis algorithm to learn spatio-temporal features combining them with deep learning
techniques such as convolutional neural networks and staking for action represen-
tation and recognition. Deep belief networks are trained with RBMs and were used
for human emotion recognition. Learning feature continuously without any labels
from the streaming video is a challenging task. Hasan and Roy-Chowdhury [40] was
addressed an unsupervised deep learning model to solve this types of problem. The
action recognition from unconstrained videos is a challenging task in most of the
20 R. Santhoshkumar and M. K. Geetha

Fig. 10 Architecture of deep convolutional neural network model

action datasets have been recorded under a controlled environment; A human action
recognition method from unconstrained video sequences was also proposed [41]
using DBNs. Unsupervised learning played a pivotal role in reviving the interests of
the researchers in deep learning environment.
Discriminative or Supervised Models The CNN is the most frequently used
model from the supervised category. The CNN is a type of deep learning model
which has shown better performance at tasks such as image classification, pattern
recognition, human emotion recognition, human action recognition and hand-written
digit classification. In hierarchical learning model, the multiple hidden layers are used
to transform the input data into output categories. Its architecture consists of three
main types of layers, such as convolutional layer, pooling layer, and fully connected
layer [42].
The objects in the images are represented and recognized by deep CNN model. The
mapping back of different layers of CNN is called as deconvolutional networks. Using
CNN approach a spatial and temporal stream for action and emotion recognition is
proposed. The combination of the two methods outperformed better results than other
methods. In supervised method the RNN is the other popular model for all the above
mentioned applications. The skeleton-based action and emotion recognition using
RNNs are developed and the five parts of human skeleton was separately fed into
five subnets [43]. The output from the subnets were combined and fed into the single
layer for final demonstration. For a training process the deep learning based model
handle a large size of video data. An outstanding accuracy level has been achieved in
many application fields. An architecture of deep convolutional neural network model
is depicted in Fig. 10.
The input videos are converted into frames and saved in separate folder as training
set and validation set. Now, the raw images are the input of first layers. The feature
difference CNN (FDCNN) consists of multiple convolutional layers, each of which
performs the function that is discussed above. The input image is of size (150 ×
150 × 3); where 3 represent colour channel. In this network the size of filter is
(5 × 5) for all layers and the filter is called as weights. The multiplying of original
pixel value with weight values is called sliding or convolving. These multiplications
are summed and produced single number is called receptive field. Each receptive
field produces a number. Finally get the feature map with size of (150 × 150 × 3).
A Study on Discrete Action Sequences Using Deep Emotional Intelligence 21

Table 7 Confusion matrix for DCNN model


Emotion Angry Fear Joy Pride Sad
Angry 98.4 1.2 0.3 0.0 0.0
Fear 0.0 98.3 0.0 0.0 1.7
Joy 0.0 0.6 99.0 0.0 0.4
Pride 0.0 0.4 0.0 98.3 1.3
Sad 0.0 0.0 0.0 0.0 100

Fig. 11 F-Score of BBIV feature with kNN, SVM, Random Forest and DCNN model

In first layer, 32 filters are applied and have 32 stacked feature maps in this stage.
Then the subsampling layer reduces the feature size of the representation with size
of (75 × 75 × 32). In second layer, 64 filters are applied and have 64 stacked feature
maps. Then the maxpooling layer is reduces the feature dimension to (37 × 37 × 64).
In the third convolutional layer, 128 numbers of filters are applied and have 128
stacked feature maps. Then the output of the maxpooling layer is reduces the feature
dimensions to (18 × 18 × 128). All max pooling layers are located with size of
(3 × 3). Finally fully connected layers with 512 hidden units are placed and the
output class have 5 neurons as per classes and shown the predicted emotions. The
confusion matrix is presented in Table 7 and the Fig. 11 shows the precision, recall,
F-score value in graphs for BBIV with KNN, RF and SVM and DCNN Model.
Figure 11 clearly shows the F-score performance for KNN, SVM and RF classifier
with BBIV and DBLHoG feature gives F-score of 89.7% and 95.3% and DCNN
model gives F-score 98.7%. From the result it is clearly indicated that the DCNN
model performs better than kNN, SVM and random forest classifiers.
22 R. Santhoshkumar and M. K. Geetha

6 Conclusion

This chapter introduced a novel approach for human emotion recognition from body
movements. Initially, a block based intensity value feature and DBLHoG features
extraction procedure was discussed for emotion recognition from pre-processed
bounding box frame. The motivation of this work is to recognize the activity and
make possible action immediately. The experimental results demonstrate that the
DCNN approach outperforms than the kNN, SVM and random forest classifiers.
Few of the promising future research directions to achieve emotional intelligence
with deep learning paradigm are discussed towards the end. The authors trust that
this attempt could afford obliging insights and significant support to the researchers
for exploring this topic. The future research direction moves on emotion recognition
of Autism Children using their movements. They can express their emotions using
heads and repeated of hand flapping.

References

1. Varghese, A.A., Cherian, J.P., Kizhakkethottam, J.J.: Overview on emotion recognition sys-
tem. In: Proceedings of International Conference on Soft Computing and Networks Security,
IEEExplore, pp. 1–5 (2015)
2. Karg, M., Samadani, A.A., Gorbet, R., Khnlenz, K., Hoey, J., Kulic, D.: Body movements for
affective expression: a survey of automatic recognition and generation. IEEE Trans. Affect.
Comput. 4(4), 341–359 (2013)
3. Piana, S., Staglian, A., Odone, F., Camurri, A.: Adaptive body gesture representation for auto-
matic emotion recognition. ACM Trans. Interact. Intell. Syst. 6(1), 1–31 (2016)
4. Glowinski, D., Dael, N., Camurri, A., Volpe, G., Mortillaro, M., Scherer, K.: Toward a minimal
representation of affective gestures. IEEE Trans. Affect. Comput. 2(2), 106–118 (2011)
5. Wang, W., Enescu, V., Sahli, H.: Adaptive real-time emotion recognition from body movements.
ACM Trans. Interact. Intell. Syst. 5(4), 1–21 (2015)
6. Fourati, N., Pelachaud, C.: Multi-level classification of emotional body expression. In: Proceed-
ings of 11th IEEE International Conference and Workshops on Automatic Face and Gesture
Recognition. IEEExplore, vol. 1, pp. 1–8 (2015)
7. Dael, N., Mortillaro, M., Scherer, K.R.: The body action and posture coding system (BAP):
development and reliability. J. Nonverbal Behav. 36(2), 97–121 (2012)
8. Stathopoulou, I.O., Tsihrintzis, G.A.: Emotion recognition from body movements and gestures.
In: Intelligent Interactive Multimedia Systems and Services, pp. 295–303. Springer, Berlin,
Heidelberg (2011)
9. Gross, M.M., Crane, E.A., Fredrickson, B.L.: Methodology for assessing bodily expression of
emotion. J. Nonverbal Behav. 34(4), 223–248 (2010)
10. Cimen, G., Ilhan, H., Capin, T., Gurcay, H.: Classification of human motion based on affective
state descriptors. Comput. Animat. Virtual Worlds 24(3–4), 355–363 (2013)
11. Zacharatos, H., Gatzoulis, C., Chrysanthou, Y.L.: Automatic emotion recognition based on
body movement analysis: a survey. IEEE Comput. Graph. Appl. 34(6), 35–45 (2014)
12. Castellano, G., Villalba, S.D., Camurri, A.: Recognising human emotions from body movement
and gesture dynamics. In: Proceedings of International Conference on Affective Computing
and Intelligent Interaction, pp. 71–82. Springer, Berlin, Heidelberg (2007)
13. Kalaiselvi Geetha, M., Palanivel, S.: Video classification and shot detection for video retrieval
applications. Int. J. Comput. Intell. Syst. 2(1), 39–50 (2009)
Other documents randomly have
different content
NOTES, OF THE BALANCE OF TRADE.

17 There is another cause, though more limited in its operation, which


checks the wrong balance of trade, to every particular nation to which
the kingdom trades. When we import more goods than we export, the
exchange turns against us, and this becomes a new encouragement to
export, as much as the charge of carriage and insurance of the money
which becomes due would amount to. For the exchange can never rise
higher than that sum.

18 It must carefully be remarked that throughout this discourse,


wherever I speak of the level of money I mean always its proportional
level to the commodities, labour, industry, and skill which is in the
several states; and I assert that where these advantages are double,
treble, quadruple to what they are in the neighbouring states, the
money infallibly will also be double, treble, quadruple. The only
circumstance that can obstruct the exactness of these proportions is the
expense of transporting the commodities from one place to another, and
this expense is sometimes unequal. Thus the corn, cattle, cheese,
butter of Derbyshire cannot draw the money of London so much as the
manufactures of London draw the money of Derbyshire. But this
objection is only a seeming one, for so far as the transport of
commodities is expensive, so far is the communication between the
places obstructed and imperfect.

19 We observed in essay Of Money, that money, when increasing,


gives encouragement to industry during the interval between the
increase of money and the rise of the prices. A good effect of this
nature may follow too from paper-credit; but it is dangerous to
precipitate matters at the risk of losing all by the failing of that credit,
as must happen upon any violent shock in public affairs.

20 There were about eight ounces of silver in a pound sterling in Harry


VII.’s time.
21 The poverty which Stanyan speaks of is only to be seen in the most
mountainous cantons, where there is no commodity to bring money;
and even there the people are not poorer than in the diocese of
Saltsburg on the one hand, or Savoy on the other.
OF THE JEALOUSY OF TRADE.
Having endeavoured to remove one species of ill-founded jealousy
which is so prevalent among commercial nations, it may not be
amiss to mention another which seems equally groundless. Nothing
is more usual, among states which have made some advances in
commerce, than to look on the progress of their neighbours with a
suspicious eye, to consider all trading states as their rivals, and to
suppose that it is impossible for any of them to flourish but at their
expense. In opposition to this narrow and malignant opinion, I will
venture to assert that the increase of riches and commerce in any
one nation, instead of hurting, commonly promotes the riches and
commerce of all its neighbours; and that a state can scarcely carry
its trade and industry very far where all the surrounding states are
buried in ignorance, sloth, and barbarism.
It is obvious that the domestic industry of a people cannot be
hurt by the greatest prosperity of their neighbours; and as this
branch of commerce is undoubtedly the most important in any
extensive kingdom, we are so far removed from all reason of
jealousy. But I go farther, and observe that where an open
communication is preserved among nations, it is impossible but the
domestic industry of every one must receive an increase from the
improvements of the others. Compare the situation of Great Britain
at present with what it was two centuries ago. All the arts, both of
agriculture and manufactures, were then extremely rude and
imperfect. Every improvement which we have since made has arisen
from our imitation of foreigners, and we ought so far to esteem it
happy that they had previously made advances in arts and ingenuity.
But this intercourse is still upheld to our great advantage.
Notwithstanding the advanced state of our manufactures, we daily
adopt in every art the inventions and improvements of our
neighbours. The commodity is {p68} first imported from abroad, to our
great discontent, while we imagine that it drains us of our money;
afterwards the art itself is gradually imported, to our visible
advantage. Yet we continue still to repine that our neighbours should
possess any art, industry, and invention, forgetting that had they not
first instructed us we should have been at present barbarians, and
did they not still continue their instructions, the arts must fall into a
state of languor, and lose that emulation and novelty which
contribute so much to their advancement.
The increase of domestic industry lays the foundation of foreign
commerce. Where a great number of commodities are raised and
perfected for the home-market there will always be found some
which can be exported with advantage. But if our neighbours have
no art nor cultivation, they cannot take them, because they will have
nothing to give in exchange. In this respect, states are in the same
condition as individuals. A single man can scarce be industrious
where all his fellow-citizens are idle. The riches of the several
members of a community contribute to increase my riches, whatever
profession I may follow. They consume the produce of my industry,
and afford me the produce of theirs in return.
Nor need any state entertain apprehensions that their neighbours
will improve to such a degree in every art and manufacture as to
have no demand from them. Nature, by giving a diversity of
geniuses, climates, and soils to different nations, has secured their
mutual intercourse and commerce, as long as they all remain
industrious and civilized. Nay, the more the arts increase in any
state, the more will be its demands from its industrious neighbours.
The inhabitants, having become opulent and skilful, desire to have
every commodity in the utmost perfection; and as they have plenty
of commodities to give in exchange, they make large importations
from every foreign country. The industry of the nations from whom
they import receives encouragement; their own is also increased by
the sale of the commodities which they give in exchange. {p69}
But what if a nation has any staple commodity, such as the
woollen manufacture is to England? Must not the interfering of their
neighbours in that manufacture be a loss to them? I answer that
when any commodity is denominated the staple of a kingdom, it is
supposed that that kingdom has some peculiar and natural
advantages for raising the commodity; and if, notwithstanding these
advantages, they lose such a manufactory, they ought to blame their
own idleness or bad government, not the industry of their
neighbours. It ought also to be considered that by the increase of
industry among the neighbouring nations the consumption of every
particular species of commodity is also increased; and though
foreign manufactures interfere with us in the market, the demand
for our product may still continue, or even increase. And even should
it diminish, ought the consequence to be esteemed so fatal? If the
spirit of industry be preserved, it may easily be diverted from one
branch to another, and the manufactures of wool, for instance, be
employed in linen, silk, iron, or other commodities for which there
appears to be a demand. We need not apprehend that all the
objects of industry will be exhausted, or that our manufacturers,
while they remain on an equal footing with those of our neighbours,
will be in danger of wanting employment; the emulation among rival
nations serves rather to keep industry alive in all of them. And any
people is happier who possess a variety of manufactures, than if
they enjoyed one single great manufacture, in which they are all
employed. Their situation is less precarious, and they will feel less
sensibly those revolutions and uncertainties to which every particular
branch of commerce will always be exposed.
The only commercial state which ought to dread the
improvements and industry of their neighbours is such a one as
Holland, which enjoying no extent of land, nor possessing any native
commodity, flourishes only by being the brokers, and factors, and
carriers of others. Such a people may naturally apprehend that as
soon as the {p70} neighbouring states come to know and pursue their
interest, they will take into their own hands the management of their
affairs, and deprive their brokers of that profit which they formerly
reaped from it. But though this consequence may naturally be
dreaded, it is very long before it takes place; and by art and industry
it may be warded off for many generations, if not wholly eluded. The
advantage of superior stocks and correspondence is so great that it
is not easily overcome; and as all the transactions increase by the
increase of industry in the neighbouring states, even a people whose
commerce stands on this precarious basis may at first reap a
considerable profit from the flourishing condition of their neighbours.
The Dutch, having mortgaged all their revenues, make not such a
figure in political transactions as formerly; but their commerce is
surely equal to what it was in the middle of the last century, when
they were reckoned among the great powers of Europe.
Were our narrow and malignant politics to meet with success, we
should reduce all our neighbouring nations to the same state of sloth
and ignorance that prevails in Morocco and the coast of Barbary. But
what would be the consequence? They could send us no
commodities, they could take none from us. Our domestic commerce
itself would languish for want of emulation, example, and
instruction; and we ourselves should soon fall into the same abject
condition to which we had reduced them. I shall therefore venture to
acknowledge that not only as a man, but as a British subject, I pray
for the flourishing commerce of Germany, Spain, Italy, and even
France itself. I am at least certain that Great Britain and all these
nations would flourish more did their sovereigns and ministers adopt
such enlarged and benevolent sentiments towards each other.
OF THE BALANCE OF POWER.
It is a question whether the idea of the balance of power be owing
entirely to modern policy, or whether the phrase only has been
invented in these latter ages. It is certain that Xenophon, in his
institution of Cyrus, represents the combination of the Asiatic powers
to have arisen from a jealousy of the increasing force of the Medes
and Persians; and though that elegant composition should be
supposed altogether a romance, this sentiment, ascribed by the
author to the Eastern princes, is at least a proof of the prevailing
notions of ancient times.
In all the politics of Greece the anxiety with regard to the balance
of power is most apparent, and is expressly pointed out to us even
by the ancient historians. Thucydides represents the league which
was formed against Athens, and which produced the Peloponnesian
war, as entirely owing to this principle. And after the decline of
Athens, when the Thebans and Lacedemonians disputed for
sovereignty, we find that the Athenians (as well as many other
republics) threw themselves always into the lighter scale, and
endeavoured to preserve the balance. They supported Thebes
against Sparta, till the great victory gained by Epaminondas at
Leuctra, after which they immediately went over to the conquered,
from generosity as they pretended, but in reality from their jealousy
of the conquerors.
Whoever will read Demosthenes’ oration for the Megalopolitans
may see the utmost refinements on this principle which ever entered
into the head of a Venetian or English speculatist; and upon the first
rise of the Macedonian power, this orator immediately discovered the
danger, sounded the alarm through all Greece, and at last assembled
that confederacy under the banners of Athens which fought the
great and decisive battle of Chæronea. {p72}
It is true the Grecian wars are regarded by historians as wars of
emulation rather than of politics, and each state seems to have had
more in view the honour of leading the rest than any well-grounded
hopes of authority and dominion. If we consider, indeed, the small
number of inhabitants in any one republic compared to the whole,
the great difficulty of forming sieges in those times, and the
extraordinary bravery and discipline of every freeman among that
noble people, we shall conclude that the balance of power was of
itself sufficiently secured in Greece, and needed not to be guarded
with that caution which may be requisite in other ages. But whether
we ascribe the shifting sides in all the Grecian republics to jealous
emulation or cautious politics, the effects were alike, and every
prevailing power was sure to meet with a confederacy against it, and
that often composed of its former friends and allies.
The same principle—call it envy or prudence—which produced
the ostracism of Athens and petalism of Syracuse, and expelled
every citizen whose fame or power overtopped the rest—the same
principle, I say, naturally discovered itself in foreign politics, and
soon raised enemies to the leading state, however moderate in the
exercise of its authority.
The Persian monarch was really, in his force, a petty prince
compared to the Grecian republics, and therefore it behoved him,
from views of safety more than from emulation, to interest himself in
their quarrels, and to support the weaker side in every contest. This
was the advice given by Alcibiades to Tissaphernes, and it prolonged
near a century the date of the Persian empire; till the neglect of it
for a moment, after the first appearance of the aspiring genius of
Philip, brought that lofty and frail edifice to the ground with a
rapidity of which there are few instances in the history of mankind.
The successors of Alexander showed an infinite jealousy of the
balance of power, a jealousy founded on true politics and prudence,
and which preserved distinct for several ages the partitions made
after the death of that famous {p73} conqueror. The fortune and
ambition of Antigonus threatened them anew with a universal
monarchy, but their combination and their victory at Ipsus saved
them; and in after times we find that as the Eastern princes
considered the Greeks and Macedonians as the only real military
force with whom they had any intercourse, they kept always a
watchful eye over that part of the world. The Ptolemies, in particular,
supported first Aratus and the Achæans, and then Cleomenes King
of Sparta, from no other view than as a counterbalance to the
Macedonian monarchs; for this is the account which Polybius gives
of the Egyptian politics.
The reason why it is supposed that the ancients were entirely
ignorant of the balance of power seems to be drawn from the
Roman history more than the Grecian, and as the transactions of the
former are generally the most familiar to us, we have thence formed
all our conclusions. It must be owned that the Romans never met
with any such general combination or confederacy against them as
might naturally be expected from their rapid conquests and declared
ambition, but were allowed peaceably to subdue their neighbours,
one after another, till they extended their dominion over the whole
known world. Not to mention the fabulous history of their Italic
wars, there was, upon Hannibal’s invasion of the Roman state, a
very remarkable crisis which ought to have called up the attention of
all civilized nations. It appeared afterwards (nor was it difficult to be
observed at the time​[22]) that this was a contest for universal
empire, and yet no prince or state seems to have been in the least
alarmed about the event or issue of the quarrel. Philip of Macedon
remained neuter till he saw the victories of Hannibal, and then most
imprudently formed an alliance with the conqueror, upon terms still
more imprudent. He stipulated that he was to assist the Carthaginian
state in their conquest of Italy, after which {p74} they engaged to
send over forces into Greece, to assist him in subduing the Grecian
commonwealths.
The Rhodean and Achæan republics are much celebrated by
ancient historians for their wisdom and sound policy; yet both of
them assisted the Romans in their wars against Philip and Antiochus.
And what may be esteemed still a stronger proof that this maxim
was not familiarly known in those ages, no ancient author has ever
remarked the imprudence of these measures, nor has even blamed
that absurd treaty above-mentioned made by Philip with the
Carthaginians. Princes and statesmen may in all ages be blinded in
their reasonings with regard to events beforehand, but it is
somewhat extraordinary that historians afterwards should not form a
sounder judgment of them.
Massinissa, Attalus, Prusias, in satisfying their private passions,
were all of them the instruments of the Roman greatness, and never
seem to have suspected that they were forging their own chains
while they advanced the conquests of their ally. A simple treaty and
agreement between Massinissa and the Carthaginians, so much
required by mutual interest, barred the Romans from all entrance
into Africa, and preserved liberty to mankind.
The only prince we meet with in the Roman history who seems
to have understood the balance of power is Hiero, King of Syracuse.
Though the ally of Rome, he sent assistance to the Carthaginians
during the war of the auxiliaries: “Esteeming it requisite,” says
Polybius, “both in order to retain his dominions in Sicily and to
preserve the Roman friendship, that Carthage should be safe; lest by
its fall the remaining power should be able, without contrast or
opposition, to execute every purpose and undertaking. And here he
acted with great wisdom and prudence; for that is never, on any
account, to be overlooked, nor ought such a force ever to be thrown
into one hand as to incapacitate the neighbouring states from
defending their rights against it.” Here is the aim of modern politics
pointed out in express terms.
In short, the maxim of preserving the balance of power is {p75}
founded so much on common sense and obvious reasoning that it is
impossible it could altogether have escaped antiquity, where we find,
in other particulars, so many marks of deep penetration and
discernment. If it was not so generally known and acknowledged as
at present, it had at least an influence on all the wiser and more
experienced princes and politicians; and indeed, even at present,
however generally known and acknowledged among speculative
reasoners, it has not, in practice, an authority much more extensive
among those who govern the world.
After the fall of the Roman Empire the form of government
established by the northern conquerors incapacitated them in a
great measure from further conquests, and long maintained each
state in its proper boundaries; but when vassalage and the feudal
militia were abolished mankind were anew alarmed by the danger of
universal monarchy, from the union of so many kingdoms and
principalities in the person of the Emperor Charles. But the power of
the house of Austria, founded on extensive but divided dominions,
and their riches, derived chiefly from mines of gold and silver, were
more likely to decay, of themselves, from internal defects, than to
overthrow all the bulwarks raised against them. In less than a
century the force of that violent and haughty race was shattered,
their opulence dissipated, their splendour eclipsed. A new power
succeeded, more formidable to the liberties of Europe, possessing all
the advantages of the former and labouring under none of its
defects, except a share of that spirit of bigotry and persecution with
which the house of Austria were so long and still are so much
infatuated.
Europe has now, for above a century, remained on the defensive
against the greatest force that ever perhaps was formed by the civil
or political combination of mankind. And such is the influence of the
maxim here treated of, that though that ambitious nation in the five
last general wars has been victorious in four,​[23] and unsuccessful
only {p76} in one,​[24] they have not much enlarged their dominions,
nor acquired a total ascendant over Europe. There remains rather
room to hope that by maintaining the resistance some time the
natural revolutions of human affairs, together with unforeseen
events and accidents, may guard us against universal monarchy, and
preserve the world from so great an evil.
In the three last of these general wars Britain has stood foremost
in the glorious struggle, and she still maintains her station as
guardian of the general liberties of Europe, and patron of mankind.
Beside her advantages of riches and situation, her people are
animated with such a national spirit, and are so fully sensible of the
inestimable blessings of their government, that we may hope their
vigour never will languish in so necessary and so just a cause. On
the contrary, if we may judge by the past, their passionate ardour
seems rather to require some moderation, and they have oftener
erred from a laudable excess than from a blameable deficiency.
In the first place, we seem to have been more possessed with
the ancient Greek spirit of jealous emulation than actuated with the
prudent views of modern politics. Our wars with France have been
begun with justice, and even, perhaps, from necessity; but have
always been too far pushed from obstinacy and passion. The same
peace which was afterwards made at Ryswick in 1697 was offered so
early as the ninety-two; that concluded at Utrecht in 1712 might
have been finished on as good conditions at Gertruytenberg in the
eight; and we might have given at Frankfort in 1743 the same terms
which we were glad to accept of at Aix-la-Chapelle in the forty-eight.
Here then we see that above half of our wars with France, and all
our public debts, are owing more to our own imprudent vehemence
than to the ambition of our neighbours.
In the second place, we are so declared in our opposition to
French power, and so alert in defence of our allies, that {p77} they
always reckon upon our force as upon their own, and expecting to
carry on war at our expense, refuse all reasonable terms of
accommodation. Habent subjectos, tanquam suos; viles, ut alienos.
All the world knows that the factious vote of the House of Commons
in the beginning of the last Parliament, with the professed humour
of the nation, made the Queen of Hungary inflexible in her terms,
and prevented that agreement with Prussia which would immediately
have restored the general tranquillity of Europe.
In the third place, we are such true combatants that, when once
engaged, we lose all concern for ourselves and our posterity, and
consider only how we may best annoy the enemy. To mortgage our
revenues at so deep a rate in wars where we are only accessories
was surely the most fatal delusion that a nation, who had any
pretension to politics and prudence, has ever yet been guilty of. That
remedy of funding—if it be a remedy and not rather a poison—
ought, in all reason, to be reserved to the last extremity, and no evil
but the greatest and most urgent should ever induce us to embrace
so dangerous an expedient.
These excesses to which we have been carried are prejudicial,
and may perhaps in time become still more prejudicial another way,
by begetting, as is usual, the opposite extreme, and rendering us
totally careless and supine with regard to the fate of Europe. The
Athenians, from the most bustling, intriguing, warlike people of
Greece, finding their error in thrusting themselves into every quarrel,
abandoned all attention to foreign affairs, and in no contest ever
took party on either side, except by their flatteries and complaisance
to the victor.
Enormous monarchies are probably destructive to human nature
—in their progress, in their continuance,​[25] and even in their
downfall, which never can be very distant from their {p78}
establishment. The military genius which aggrandized the monarchy
soon leaves the court, the capital, and the centre of such a
government; while the wars are carried on at a great distance, and
interest so small a part of the state. The ancient nobility, whose
affections attach them to their sovereign, live all at court; and never
will accept of military employments which would carry them to
remote and barbarous frontiers, where they are distant both from
their pleasures and their fortune. The arms of the state must
therefore be trusted to mercenary strangers, without zeal, without
attachment, without honour, ready on every occasion to turn them
against the prince, and join each desperate malcontent who offers
pay and plunder. This is the necessary progress of human affairs;
thus human nature checks itself in its airy elevations, thus ambition
blindly labours for the destruction of the conqueror, of his family, and
of everything near and dear to him. The Bourbons, trusting to the
support of their brave, faithful, and affectionate nobility, would push
their advantage without reserve or limitation. These, while fired with
glory and emulation, can bear the fatigues and dangers of war; but
never would submit to languish in the garrisons of Hungary or
Lithuania, forgot at court, and sacrificed to the intrigues of every
minion or mistress who approaches the prince. The troops are filled
with Cravates and Tartars, Hussars and Cossacks, intermingled
perhaps with a few soldiers of fortune from the better provinces;
and the melancholy fate of the Roman emperors, from the same
cause, is renewed over and over again till the final dissolution of the
monarchy.
NOTES, OF THE BALANCE OF POWER.

22 It was observed by some, as appears from the speech of Agelaus of


Naupactum, in the general congress of Greece. See Polyb., lib. 5, cap.
104.

23 Those concluded by the Peace of the Pyrenees, Nimeguen, Ryswick,


and Aix-la-Chapelle.

24 That concluded by the Peace of Utrecht.

25 If the Roman Empire was of advantage, it could only proceed from


this, that mankind were generally in a very disorderly, uncivilized
condition before its establishment.
OF TAXES.
There is a maxim that prevails among those whom in this country
we call “ways and means” men, and who are denominated financiers
and maltotiers in France, that every new tax creates a new ability in
the subject to bear it, and {p79} that each increase of public burdens
increases proportionably the industry of the people. This maxim is of
such a nature as is most likely to be extremely abused, and is so
much the more dangerous, as its truth cannot be altogether denied;
but it must be owned, when kept within certain bounds, to have
some foundation in reason and experience.
When a tax is laid upon commodities which are consumed by the
common people, the necessary consequence may seem to be that
either the poor must retrench something from their way of living, or
raise their wages so as to make the burden of the tax fall entirely
upon the rich. But there is a third consequence which very often
follows upon taxes—viz., that the poor increase their industry,
perform more work, and live as well as before without demanding
more for their labour. Where taxes are moderate, are laid on
gradually, and affect not the necessaries of life, this consequence
naturally follows; and it is certain that such difficulties often serve to
excite the industry of a people, and render them more opulent and
laborious than others who enjoy the greatest advantages. For we
may observe, as a parallel instance, that the most commercial
nations have not always possessed the greatest extent of fertile
land; but, on the contrary, that they have laboured under many
natural disadvantages. Tyre, Athens, Carthage, Rhodes, Genoa,
Venice, Holland are strong examples to this purpose; and in all
history we find only three instances of large and fertile countries
which have possessed much trade—the Netherlands, England, and
France. The two former seem to have been allured by the
advantages of their maritime situation, and the necessity they lay
under of frequenting foreign ports in order to procure what their
own climate refused them; and as to France, trade has come very
late into the kingdom, and seems to have been the effect of
reflection and observation in an ingenious and enterprising people,
who remarked the immense riches acquired by such of the
neighbouring nations as cultivated navigation and commerce.
The places mentioned by Cicero as possessed of the {p80} greatest
commerce of his time are Alexandria, Colchos, Tyre, Sidon, Andros,
Cyprus, Pamphylia, Lycia, Rhodes, Chios, Byzantium, Lesbos,
Smyrna, Miletum, Coos. All these, except Alexandria, were either
small islands or narrow territories; and that city owed its trade
entirely to the happiness of its situation.
Since, therefore, some natural necessities or disadvantages may
be thought favourable to industries, why may not artificial burdens
have the same effect? Sir William Temple,​[26] we may observe,
ascribes the industry of the Dutch entirely to necessity, proceeding
from their natural disadvantages; and illustrates his doctrine by a
very striking comparison with Ireland, “where,” says he, “by the
largeness and plenty of the soil, and scarcity of people, all things
necessary to life are so cheap that an industrious man by two days’
labour may gain enough to feed him the rest of the week. Which I
take to be a very plain ground of the laziness attributed to the
people. For men naturally prefer ease before labour, and will not
take pains if they can live idle; though when, by necessity, they have
been inured to it, they cannot leave it, being grown a custom
necessary to their health, and to their very entertainment. Nor
perhaps is the change harder from constant ease to labour than
from constant labour to ease.” After which the author proceeds to
confirm his doctrine by enumerating as above the places where
trade has most flourished in ancient and modern times, and which
are commonly observed by such narrow, confined territories as
beget a necessity for industry.
It is always observed in years of scarcity, if it be not extreme,
that the poor labour more and really live better than in years of
great plenty, when they indulge themselves in idleness and riot. I
have been told by a considerable manufacturer that in the year
1740, when bread and provisions of all kinds were very dear, his
workmen not only made a shift to live, but paid debts which they
had {p81} contracted in former years that were much more favourable
and abundant.
This doctrine, therefore, with regard to taxes may be admitted to
some degree, but beware of the abuse. Exorbitant taxes, like
extreme necessity, destroy industry by producing despair; and even
before they reach this pitch they raise the wages of the labourer and
manufacturer, and heighten the price of all commodities. An
attentive, disinterested legislature will observe the point when the
emolument ceases and the prejudice begins; but as the contrary
character is much more common, it is to be feared that taxes all
over Europe are multiplying to such a degree as will entirely crush all
art and industry; though perhaps their first increase, together with
circumstances, might have contributed to the growth of these
advantages.
The best taxes are such as are levied upon consumptions,
especially those of luxury, because such taxes are less felt by the
people. They seem, in some measure, voluntary, since a man may
choose how far he will use the commodity which is taxed: they are
paid gradually and insensibly, and being confounded with the natural
price of the commodity, they are scarcely perceived by the
consumers. Their only disadvantage is that they are expensive in the
levying.
Taxes upon possessions are levied without expense, but have
every other disadvantage. Most states, however, are obliged to have
recourse to them, in order to supply the deficiencies of the other.
But the most pernicious of all taxes are those which are arbitrary.
They are commonly converted by their management into
punishments on industry; and also by their unavoidable inequality
are more grievous than by the real burden which they impose. It is
surprising, therefore, to see them have place among any civilized
people.
In general, all poll-taxes, even when not arbitrary—which they
commonly are—may be esteemed dangerous; because it is so easy
for the sovereign to add a little more and a little more to the sum
demanded, that these taxes are apt to become altogether oppressive
and intolerable. On the {p82} other hand, a duty upon commodities
checks itself, and a prince will soon find that an increase of the
impost is no increase of his revenue. It is not easy, therefore, for a
people to be altogether ruined by such taxes.
Historians inform us that one of the chief causes of the
destruction of the Roman state was the alteration which Constantine
introduced into the finances, by substituting a universal poll-tax in
lieu of almost all the tithes, customs, and excises which formerly
composed the revenue of the empire. The people in all the provinces
were so grinded and oppressed by the publicans that they were glad
to take refuge under the conquering arms of the barbarians, whose
dominion, as they had fewer necessities and less art, was found
preferable to the refined tyranny of the Romans.
There is a prevailing opinion that all taxes, however levied, fall
upon the land at last. Such an opinion may be useful in Britain, by
checking the landed gentlemen, in whose hands our legislature is
chiefly lodged, and making them preserve great regard for trade and
industry; but I must confess that this principle, though first
advanced by a celebrated writer, has so little appearance of reason
that were it not for his authority it had never been received by
anybody. Every man, to be sure, is desirous of pushing off from
himself the burden of any tax which is imposed, and laying it upon
others; but as every man has the same inclination, and is upon the
defensive, no set of men can be supposed to prevail altogether in
this contest. And why the landed gentleman should be the victim of
the whole, and should not be able to defend himself as well as
others are, I cannot readily imagine. All tradesmen, indeed, would
willingly prey upon him and divide him among them if they could;
but this inclination they always have, though no taxes were levied;
and the same methods by which he guards against the imposition of
tradesmen before taxes will serve him afterwards, and make them
share the burden with him. No labour in any commodities that are
exported can be very considerably raised in the price without losing
the foreign market; and as some part of almost every {p83}
manufactory is exported, this circumstance keeps the price of most
species of labour nearly the same after the imposition of taxes. I
may add that it has this effect upon the whole, for were any kind of
labour paid beyond its proportion all hands would flock to it, and
would soon sink it to a level with the rest.
I shall conclude this subject with observing that we have with
regard to taxes an instance of what frequently happens in political
institutions, that the consequence of things are diametrically
opposite to what we should expect on the first appearance. It is
regarded as a fundamental maxim of the Turkish Government that
the Grand Seignior, though absolute master of the lives and fortunes
of each individual, has no authority to impose a new tax; and every
Ottoman prince who has made such an attempt either has been
obliged to retract, or has found the fatal effects of his perseverance.
One would imagine that this prejudice or established opinion were
the firmest barrier in the world against oppression, yet it is certain
that its effect is quite contrary. The emperor, having no regular
method of increasing his revenue, must allow all the pashas and
governors to oppress and abuse the subjects, and these he squeezes
after their return from their government; whereas, if he could
impose a new tax, like our European princes, his interest would so
far be united with that of his people that he would immediately feel
the bad effects of these disorderly levies of money, and would find
that a pound raised by general imposition would have less pernicious
effects than a shilling taken in so unequal and arbitrary a manner.
NOTE, OF TAXES.

26 Account of the Netherlands, chap. vi.


OF PUBLIC CREDIT.
It appears to have been the common practice of antiquity to make
provision in times of peace for the necessities of war, and to hoard
up treasures beforehand as the instruments either of conquest or
defence, without trusting to {p84} extraordinary imposts, much less to
borrowing, in times of disorder and confusion. Besides the immense
sums above mentioned​[27] which were amassed by Athens, and by
the Ptolemies and other successors of Alexander, we learn from Plato
that the frugal Lacedemonians had also collected a great treasure;
and Arrian and Plutarch​[28] specify the riches which Alexander got
possession of on the conquest of Susa and Ecbatana, and which
were reserved, some of them, from the time of Cyrus. If I remember
right, the Scripture also mentions the treasure of Hezekiah and the
Jewish princes, as profane history does that of Philip and Perseus,
kings of Macedon. The ancient republics of Gaul had commonly large
sums in reserve. Every one knows the treasure seized in Rome by
Julius Cæsar during the civil wars, and we find afterwards that the
wiser emperors, Augustus, Tiberius, Vespasian, Severus, etc., always
discovered the prudent foresight of saving great sums against any
public exigency.
On the contrary, our modern expedient, which has become very
general, is to mortgage the public revenues, and to trust that
posterity during peace will pay off the encumbrances contracted
during the preceding war; and they, having before their eyes so
good an example of their wise fathers, have the same prudent
reliance on their posterity, who at last, from necessity more than
choice, are obliged to place the same confidence in a new posterity.
But not to waste time in declaiming against a practice which appears
ruinous beyond the evidence of a hundred demonstrations, it seems
pretty apparent that the ancient maxims are in this respect much
more prudent than the modern; even though the latter had been
confined within some reasonable bounds, and had ever, in any one
instance, been attended with such frugality in time of peace as to
discharge the debts incurred by an expensive war. For why should
the case be so very different between the public and an individual as
to make {p85} us establish such different maxims of conduct for each?
If the funds of the former be greater, its necessary expenses are
proportionably larger; if its resources be more numerous, they are
not infinite; and as its frame should be calculated for a much longer
duration than the date of a single life, or even of a family, it should
embrace maxims, large, durable, and generous, agreeable to the
supposed extent of its existence. To trust to chances and temporary
expedients is indeed what the necessity of human affairs frequently
reduces it to, but whoever voluntarily depend on such resources
have not necessity but their own folly to accuse for their misfortunes
when any such befall them.
If the abuses of treasures be dangerous, either by engaging the
state in rash enterprises or making it neglect military discipline in
confidence of its riches, the abuses of mortgaging are more certain
and inevitable—poverty, impotence, and subjection to foreign
powers.
According to modern policy, war is attended with every
destructive circumstance: loss of men, increase of taxes, decay of
commerce, dissipation of money, devastation by sea and land.
According to ancient maxims, the opening of the public treasure, as
it produced an uncommon affluence of gold and silver, served as a
temporary encouragement to industry, and atoned in some degree
for the inevitable calamities of war.
What then shall we say to the new paradox, that public
encumbrances are, of themselves, advantageous, independent of the
necessity of contracting them; and that any state, even though it
were not pressed by a foreign enemy, could not possibly have
embraced a wiser expedient for promoting commerce and riches
than to create funds, and debts, and taxes without limitation?
Discourses such as these might naturally have passed for trials of wit
among rhetoricians, like the panegyrics on folly and a fever, on
Busiris and Nero, had we not seen such absurd maxims patronized
by great ministers and by a whole party among us; and these
puzzling arguments (for they deserve not the name of specious),
though they could not be the foundation of Lord {p86} Orford’s
conduct, for he had more sense, served at least to keep his partisans
in countenance and perplex the understanding of the nation.
Let us examine the consequences of public debts, both in our
domestic management by their influence on commerce and industry,
and in our foreign transactions by their effect on wars and
negotiations.
There is a word which is here in the mouth of everybody, and
which I find has also got abroad and is much employed by foreign
writers​[29] in imitation of the English—and this is “circulation.” This
word serves as an account of everything, and though I confess that
I have sought for its meaning in the present subject ever since I was
a schoolboy, I have never yet been able to discover it. What possible
advantage is there which the nation can reap by the easy
transference of stock from hand to hand? Or is there any parallel to
be drawn from the circulation of other commodities to that of
chequer notes and India bonds? Where a manufacturer has a quick
sale of his goods to the merchant, the merchant to the shopkeeper,
the shopkeeper to his customers, this enlivens industry and gives
new encouragement to the first dealer or the manufacturer and all
his tradesmen, and makes them produce more and better
commodities of the same species. A stagnation is here pernicious,
wherever it happens, because it operates backwards, and stops or
benumbs the industrious hand in its production of what is useful to
human life. But what production we owe to Change-alley, or even
what consumption, except that of coffee, and pen, ink, and paper, I
have not yet learned; nor can one foresee the loss or decay of any
one beneficial commerce or commodity, though that place and all its
inhabitants were for ever buried in the ocean.
But though this term has never been explained by those who
insist so much on the advantages that result from a circulation, there
seems, however, to be some benefit of a similar kind arising from
our encumbrances—as, indeed, {p87} what human evil is there which
is not attended with some advantage? This we shall endeavour to
explain, that we may estimate the weight which we ought to allow
it.
Public securities are with us become a kind of money, and pass
as readily at the current price as gold or silver. Wherever any
profitable undertaking offers itself, however expensive, there are
never wanting hands enough to embrace it; nor need a trader who
has sums in the public stocks fear to launch out into the most
extensive trade, since he is possessed of funds which will answer the
most sudden demand that can be made upon him. No merchant
thinks it necessary to keep by him any considerable cash. Bank-
notes or India bonds, especially the latter, serve all the same
purposes; because he can dispose of them or pledge them to a
banker in a quarter of an hour; and at the same time they are not
idle, even when in his escritoire, but bring him in a constant
revenue. In short, our national debts furnish merchants with a
species of money that is continually multiplying in their hands, and
produces sure gain besides the profits of their commerce. This must
enable them to trade upon less profit. The small profit of the
merchant renders the commodity cheaper, causes a greater
consumption, quickens the labour of the common people, and helps
to spread arts and industry through the whole society.
There are also, we may observe, in England and in all states
which have both commerce and public debts, a set of men who are
half merchants, half stock-holders, and may be supposed willing to
trade for small profits; because commerce is not their principal or
sole support, and their revenues in the funds are a sure resource for
themselves and their families. Were there no funds great merchants
would have no expedient for realizing or securing any part of their
profit but by making purchases of land, and land has many
disadvantages in comparison of funds. Requiring more care and
inspection, it divides the time and attention of the merchant; upon
any tempting offer or extraordinary accident in trade, it is not so
easily converted into money; and as it {p88} attracts too much, both
by the many natural pleasures it affords and the authority it gives, it
soon converts the citizen into the country gentleman. More men,
therefore, with large stocks and incomes, may naturally be supposed
to continue in trade where there are public debts; and this, it must
be owned, is of some advantage to commerce by diminishing its
profits, promoting circulation, and encouraging industry.
But, in opposition to these two favourable circumstances,
perhaps of no very great importance, weigh the many disadvantages
which attend our public debts in the whole interior economy of the
state; you will find no comparison between the ill and the good
which result from them.
First, it is certain that national debts cause a mighty confluence
of people and riches to the capital, by the great sums which are
levied in the provinces to pay the interest of those debts; and
perhaps, too, by the advantages in trade above mentioned, which
they give the merchants in the capital above the rest of the
kingdom. The question is, whether, in our case, it be for the public
interest that so many privileges should be conferred on London,
which has already arrived at such an enormous size and seems still
increasing? Some men are apprehensive of the consequences. For
my part, I cannot forbear thinking that though the head is
undoubtedly too big for the body, yet that great city is so happily
situated that its excessive bulk causes less inconvenience than even
a smaller capital to a greater kingdom. There is more difference
between the prices of all provisions in Paris and Languedoc than
between those in London and Yorkshire.
Secondly, public stocks, being a kind of paper-credit, have all the
disadvantages attending that species of money. They banish gold
and silver from the most considerable commerce of the state, reduce
them to common circulation, and by that means render all provisions
and labour dearer than otherwise they would be.
Thirdly, the taxes which are levied to pay the interests of these
debts are apt to be a check upon industry, to {p89} heighten the price
of labour, and to be an oppression on the poorer sort.
Fourthly, as foreigners possess a share of our national funds,
they render the public in a manner tributary to them, and may in
time occasion the transport of our people and our industry.
Fifthly, the greatest part of public stock being always in the
hands of idle people, who live on their revenue, our funds give great
encouragement to a useless and inactive life.
But though the injury which arises to commerce and industry
from our public funds will appear, upon balancing the whole, very
considerable, it is trivial in comparison of the prejudice which results
to the state considered as a body politic, which must support itself in
the society of nations, and have various transactions with other
states, in wars and negotiations. The ill there is pure and unmixed,
without any favourable circumstance to atone for it, and it is an ill
too of a nature the highest and most important.
We have, indeed, been told that the public is no weaker upon
account of its debts, since they are mostly due among ourselves,
and bring as much property to one as they take from another. It is
like transferring money from the right hand to the left, which leaves
the person neither richer nor poorer than before. Such loose
reasonings and specious comparisons will always pass where we
judge not upon principles. I ask, is it possible, in the nature of
things, to overburden a nation with taxes, even where the sovereign
resides among them? The very doubt seems extravagant, since it is
requisite in every commonwealth that there be a certain proportion
observed between the laborious and the idle part of it. But if all our
present taxes be mortgaged, must we not invent new ones? and
may not this matter be carried to a length that is ruinous and
destructive?
In every nation there are always some methods of levying money
more easy than others, agreeable to the way of living of the people
and the commodities they make use of. In Britain the excises upon
malt and beer afford a very large {p90} revenue, because the
operations of malting and brewing are very tedious, and are
impossible to be concealed; and at the same time, these
commodities are not so absolutely necessary to life as that the
raising their price would very much affect the poorer sort. These
taxes being all mortgaged, what difficulty to find new ones! what
vexation and ruin of the poor!
Duties upon consumptions are more equal and easy than those
upon possessions. What a loss to the public that the former are all
exhausted, and that we must have recourse to the more grievous
method of levying taxes!
Were all the proprietors of land only stewards to the public, must
not necessity force them to practise all the arts of oppression used
by stewards, where the absence or negligence of the proprietor
render them secure against inquiry?
It will scarce be asserted that no bounds ought ever to be set to
national debts, and that the public would be no weaker were twelve
or fifteen shillings in the pound land-tax mortgaged, with the
present customs and excises. There is something therefore in the
case beside the mere transferring of property from one hand to
another. In 500 years the posterity of those now in the coaches and
of those upon the boxes will probably have changed places, without
affecting the public by these revolutions.
Suppose the public once fairly brought to that condition to which
it is hastening with such amazing rapidity; suppose the land to be
taxed eighteen or nineteen shillings in the pound (for it can never
bear the whole twenty); suppose all the excises and customs to be
screwed up to the outmost which the nation can bear, without
entirely losing its commerce and industry; and suppose that all those
funds are mortgaged to perpetuity, and that the invention and wit of
all our projectors can find no new imposition which may serve as the
foundation of a new loan; and let us consider the necessary
consequences of this situation. Though the imperfect state of our
political knowledge and the narrow capacities of men make it
difficult to foretell the effects {p91} which will result from any untried
measure, the seeds of ruin are here scattered with such profusion as
not to escape the eye of the most careless observer.
In this unnatural state of society, the only persons who possess
any revenue beyond the immediate effects of their industry are the
stockholders, who draw almost all the rent of the land and houses,
besides the produce of all the customs and excises. These are men
who have no connections in the state, who can enjoy their revenue
in any part of the world in which they choose to reside, who will
naturally bury themselves in the capital, or in great cities, and who
will sink into the lethargy of a stupid and pampered luxury, without
spirit, ambition, or enjoyment. Adieu to all ideas of nobility, gentry,
and family. The stocks can be transferred in an instant, and being in
such a fluctuating state, will seldom be transmitted during three
generations from father to son. Or were they to remain ever so long
in one family, they convey no hereditary authority or credit to the
possessors; and by this means, the several ranks of men, which
form a kind of independent magistracy in a state, instituted by the
hand of nature, are entirely lost, and every man in authority derives
his influence from the commission alone of the sovereign. No
expedient remains for preventing or suppressing insurrections but
mercenary armies; no expedient at all remains for resisting tyranny;
elections are swayed by bribery and corruption alone; and the
middle power between king and people being totally removed, a
horrible despotism must infallibly prevail. The landholders, despised
for their poverty and hated for their oppressions, will be utterly
unable to make any opposition to it.
Though a resolution should be formed by the legislature never to
impose any tax which hurts commerce and discourages industry, it
will be impossible for men, in subjects of such extreme delicacy, to
reason so justly as never to be mistaken, or amidst difficulties so
urgent, never to be seduced from their resolution. The continual
fluctuations in commerce require continual alterations in the nature
of the taxes, which exposes the legislature every moment to {p92} the
danger both of wilful and involuntary error; and any great blow
given to trade, whether by injudicious taxes or by other accidents,
throws the whole system of the government into confusion.
But what expedient is the public now to fall upon, even
supposing trade to continue in the most flourishing condition, to

You might also like