HAND WRITTEN PATTERN RECOGNITION USING CONVOLUTIONAL

NEURAL NETWORK
A Senior Design Project Report
Submitted to
Siksha 'O' Anusandhan University, Bhubaneswar
Department of Electrical & Electronics Engineering
Institute of Technical Education & Research, Bhubaneswar
In partial fulfillment of the requirement for the
AWARD OF DEGREE OF BACHELOR OF TECHNOLOGY IN
Electrical & Electronics Engineering
Submitted By
Name:- Shibani Khatua
(1641014201)
Shivani Kumari(1641014027)

Group -7-Section-B(8th Semester)

DEPARTMENT OF ELECTRICAL & ELECTRONICS

ENGINEERING
ITER , SIKSHA 'O' ANUSANDHAN UNIVERSITY
(Declared u/s. 3 of the UGC Act. 1956)
Bhubaneswar-751030
JUNE-2020

1
 CERTIFICATE

This is to certify that the project report entitled “Hand written pattern recognition using
convolutional neural network-” being submitted by Shibani Khatua& Shivani Kumari to the
Institute of Technical Education and Research, Siksha ‘O’ Anusandhan Deemed to be
University, Bhubaneswar for the partial fulfillment for the degree of Bachelor of Technology
inElectrical & Electronics Engineeringis a record of original bonafide work carried out by them
under our supervision and guidance. The project work, in our opinion, has reached the requisite
standard fulfilling the requirements for the degree of Bachelor ofTechnology.

TheresultscontainedinthisreporthavenotbeensubmittedinpartorfulltoanyotherUniversity or
Institute for the award of any degree ordiploma.

(InternalDesignPanel-EEE) (Internal DesignPanel-EEE)

(InternalDesignPanel-EEE) (Internal DesignPanel-EEE)

2
 ACKNOWLEDGEMENT

We have taken efforts in this project. However, it would not have been possible without the kind
support and help of many individuals of our institution. Through this project work we have learnt a lot
about designing processes and considerations, which will help us infuture.
We would like to express our special thanks of gratitude to Mr. Mrutyunjaya Sahanifor his guidance
and support in completion of the project.
We are also very thankful to Mr. Mrutyunjaya Sahanifor his guidance in doing the calculation and
designing process throughout the project completion span.
We are grateful to Dr. N.Nayak, HOD Department of Electrical & Electronics Engineering for
allowing us to use all the facilities available in the college laboratories and library along with other
facilities, without which the completion of the project is impossible.

3
 DECLARATION

We announce this composed accommodation speaks to our thoughts in our own words and
whereothers thoughts or words have been incorporated, we have enough referredto and referenced
the firstsources. We additionally pronounce that we have clung to all standards of scholastic
trustworthiness and respectability and have not distorted or manufactured or misrepresented any
thought/truth/source in our accommodation.
We comprehend that any infringement of the above will be cause for disciplinary activity by the
University and can likewise bring out reformatory activity from the sources which have therefore
not been appropriately referred to or from whom legitimate consent has not been taken when
required.

Name : Shibani Khatua(Registration No -1641014201)

Name: Shivani Kumari (Registration No:- 1641014027)

Place:Bhubaneswar

Date:-- 06.2020

4
 REPORT APPROVAL

This project report entitled “Hand written pattern recognition using convolutional neural network“
By Mr. Mrutyunjaya Sahani group-11 section-B is approved for the degree of the Bachelor of Technology in
Electrical & Electronics Engineering

Examiners

Supervisor

Mrutyunjaya Sahani

Chairman

Place:Bhubaneswar

Date: June2020

5
 ABSTRACT

Handwritten character recognition has been one of the active and challenging

research areas in the field of image processing and pattern recognition. A neural

network is a feed forward neural network used for classification and recognition of

handwritten characters using numerous steps. It involves numerous applications

which are reading aid for blind, bank cheques and conversion of any handwritten

document into structural text form. An attempt is made to recognize handwritten

characters for English alphabets without feature extraction using multilayer Feed

Forward neural network .Each character data set contains 26 alphabets .Fifty different

character data sets are used for training the neural network. In the proposed system,

each character has been resized into 30x20 pixels, which is directly subjected to

training. That is, each resized character has 600 pixels and these pixels are taken as

features for training the neural network . The results show that the proposed system

yields good recognition rates which are comparable to that of feature extraction based

schemes for handwritten character recognition.

6
 TABLE OF CONTENTS

CONTENT Pages

1 Chapter-1 : Introduction 11-16

2 Chapter-2 : Customer Needs Recognition 17-21
3 Chapter-3 : Function decomposition 22-23
4 Chapter-4 : Product Teardown and Engg. Specifications 24
5 Chapter-5 : Product Architecture 25-28
6 Chapter-6 : Concept generation 29-30
7 Chapter-7 :Concept selection 31-32
8 Chapter-8 :Product Embodiment 33-36
9 Chapter-9 :Analytical and Numerical Model solution 37-43
10 Chapter-10 :Bill of materials and cost of product 44
11 Chapter-11 :Conclusions 47

12 Plagiarism 49

7
 LIST OF FIGURES

Fig no. Description of figure Page no.

22
1 BLOCK DIAGRAM OF TRADITIONAL
METHOD
BLOCK DIAGRAM OF MODERN 23
2 METHOD

BLOCK DIAGRAM OF 23
3 CONVOLUTIONAL NEURAL
NETWORK
BLOCK DIAGRAM OF WHOLE SYSTEM 24
4

. BLOCK DIAGRAM OF CONTROL 26

5 FLOW DIAGRAM

OVERVIEW OF DIGITS TAKEN UNDER 32

6 EXPEIMENT

OVERVIEW OF CONVOLUTIONAL 37
7 NEURAL NETWORK
ARCHITECTURE

OVERVIEW OF CONVOLUTIONAL
8 NEURAL NETWORK 39
FRAMEWORK
OVERVIEW OF CONVOLUTIONAL 39
9 NEURAL NETWORK

ARCHITURE OF THE 41
10 CONVOLUTIONAL NEURAL
NETWORK
CONTROL FLOW DIAGRAM 42
11

8
 OVERVIEW OF ENCODING AND 44
12 DECODING LAYER

OVERVIEW OF HIDDEN LAYERS 45

13 SHOWING PRIMARY AND
SECONDARY FEATURE
OVERVIEW OF SOFTMAX CALSSIFIER LAYER 45
14

15 OVERVIEW OF PROPOSED NETWORK 46

STRUCTURE

16 OVERVIEW OF CONVOLUTIONAL 46
NETWORK STRUCTURE

17 OVERVIEW OF FLOW DIAGRAM FOR 47

CHARACTER RECOGNITION

18 OVERVIEW OF FIRST SPARSE 48

NETRWORK LAYER STRUCTURE

SECOND SPARSE NETWORK

19 OVERVIEW OF SECOND SPARSE 48

NETRWORK LAYER STRUCTURE

20 OVERVIEW OF SOFTMAX LAYER 49

STRUCTURE

21 OVERVIEW OF DIFFERENT NETWORK 49

LAYER STRUCTURE

22 OVERVIEW OF DIFFERENT NEURAL 50

NETWORK COMBINATIONAL
STRUCTURE

9
PROBLEM STATEMENT

The problem revolves around the data storage of handwritten and paper documents to

a memory disk where it can be used and edited as per the requirement. These days, the

scanning of documents by scanner does not produce accurate results in order to store

in a digital format and for this reason it takes a lot more time to type each single word

manually for the digital storage. Thus, one needs to find a different software , system

or algorithm which can detect these handwritten documents and can be converted to a

digital format along with storage can take place easily in memory devices for further

modifications .This solution could also help in maintaining the data and saving it for

further uses as the requirements of the user.

Chapter 1
10
 Introduction

1.1 The broad statement of design problem

1.1 Student outcomes and Bloom’s taxonomy of learning levels. In

total here are 11 outcomes which are expected from students by

the time of graduation. The outcomes termed as [a-k] as

described by table 1.1

Table 1.1: Student outcomes

Description
Outcome
A An ability to apply knowledge of mathematics, science and
engineering
B An ability to design and conduct experiments, as well as to
analyze and interpret data
C An ability to design a system, component or process to meet
desired needs within realistic constraints such as economic,
environmental, social, political, safety, and sustainability
D An ability to function on multidisciplinary teams
E An ability to identify, formulate and solve engineering problems
F An understanding of professional and ethical responsibility
G An ability to communicate effictively
H The broad education necessary to understand the impact of engineering
solution in a globe, economic, environmental and societal
context
I A recognition of the need for and an ability to engage in life- long
learning
J A knowledge of contemporary issues
K An ability to use the techniques, skills and modern engineering tools
necessary for engineering practice.

The six levels of learning according to bloom's taxonomy

11
has been given in table as multi-tiered model.

Table 1.2

Level Name Description

L-1 Knowledge Retrieving, recognizing and recalling relevant

knowledge from long term memory

L-2 Constructing meaning from oral, written and

Comprehension graphic messages through interpreting,
exemplifying, classifying, summarizing,
inferring, comparing and explaining
L-3 Application Carrying out or using a procedure through
executing or implementing

L-4 Analysis Breaking material into constituent parts,

determining how the parts relate to one
another and to an overall structure or
purpose through differentiating, organizing
and attributing
L-5 Evaluation Making judgments based on criteria
and standards through checking and
critiquing
L-6 Creation Putting elements together to form a coherent
or functional whole recognizing elements
into a new pattern or structure

12
 1.3 Course results

The Course had been organized to such an extent that the under studies get the chance

to put his/her structure and assembling aptitude into utilization alongside the chance to

feature the capacity to get proficient and moral duties..

1.3 In the table beneath we have referenced the course results.

Outcom Description
Number
CO-1 Understand concepts of product design process.
CO-2 Work effectively in team to achieve a common goal.
CO-3 Gathering and analyzing customer needs for design of product.
CO-4 Decompose a product design problem into simpler sub problems for
analysis.
CO-5 Evaluate the quality, value and performance of
the product through benchmarking.
CO-6 Develop possible alternatives and selecting the best one for solving
the product function.
CO-7 Develop model and analyze it by various methods.
CO-8 Design for manufacturing and assembly a product by effective
methodology to reduce product cost.
CO-9 Develop the physical prototype and experiment it using
experimental techniques.
CO-10 Use design resources such as professional journals, trade journals,
catalogs, and the internet in project design for lifelong learning.
CO-11 Understand the global issues and social responsibility in product design.
CO-12 Discuss various socio-economic issues related to the developed
product.
CO-13 Understand professional code of ethics.
CO-14 Communicate effectively.

13
 Table 1.3:- Course Outcomes

checkpoint A B C D E F G H I J K
1 Customer √ √
needs
recognitio
n
2 Function √ √
decompositio
n
3 Engineering √ √
specification
4 Product √ √
architecture
5 Concept √ √
generation
6 Concept √ √ √ √ √ √
selection
7 Product √ √
embodiment
8 Product √ √ √ √
metric model
9 DFM, DFA, √ √ √ √ √ √
DFE
10 Analysis and √ √ √ √ √
numerical
solution
Table 1,4:- Design checkpoint and student outcome

1.5:-Timeline for design checkpoints

Checkpoints Set time line

1 Customer needs recognition 7 days
2 Function decomposition 2 day
3 Engineering specification 1 day
4 Product architecture 2 days
5 Concept generation 5 days
6 Concept selection 3 days
7 Product embodiment 1 day
10 Analytical and numerical solution 2 days

14
INTRODUCTION:-.

Hand written pattern are trouble-free to understand by humans being, as they have the

capability to learn. This capability has been feed to the machinery by artificial

intelligence and machine knowledge . And the field which deals with these pattern are

known as OCR ( optical character recognition ) . Pattern identification is an art of

detecting, segmenting and identifying font from a picture .Its purpose is to transfer the

handwriting of the human being into machine programmed type. An ultimate purpose

of hand written character detection is to copy the human reading capabilities so that

the computer can translate , recognize, correct, as he work as human do with

manuscript Handwritten recognition has been one of the most interesting and

demanding in research areas in field of image processing and pattern recognition in the

modern days .It contribute enormously in the development of automation process and

improve the limit between human being and device in various application. Numerous

study works have been focusing on new techniques and methods that would decrease

the processing time while providing superior recognition accuracy. Pattern

recognition is mainly of two types online and offline. Online pattern recognition,

information is capture throughout the writing process with the aid of a particular pen

on electronic surface. Offline identification involve scanning of information from a

document. The scanned information are massive in size and many processing

applications as searching for a content, editing, maintenance are either rigid or not

possible.

The online mode of recognition is typically used to identify only human handwriting.

In this the handwriting is capture and store in digital form through different way.

Page 15
Generally, a particular pen is used in combination with an electronic surface. As the

pen moves across the surface, the two- dimensional coordinates of successive points

are represented as a function of time and are stored in order. The online handwriting

recognition has great potential to improve user and computer communication. Various

applications including mail sorting, bank processing, document reading and postal

address recognition require offline handwriting recognition systems. As a result,

the off-line handwriting recognition continues to be an active area for research towards

exploring the newer technique that would progress in recognition accuracy .

Page 16
 Chapter 2

Customer Need Recognition

Task Distribution and Monitoring

Table 2.1: Work Distribution

Work done by each member of Team of EEE sec-“C”

Team members Task(s) to be performed Task assigned Task completed
date on
SHIBANI COLLECT DATA AND 12.5.2020 18.5.2020
KHATUA REPORT

SHIVANI KUMARI COLLECT DATA AND 12.5.2020 18.5.2020

REPORT

Direct Need- Direct use of pattern recognizer is in the field of medical science like biological and

biomedical imaging. Pattern recognition is used to extract meaningful features from given image sample

and is used in computer vision for various application.

Latent Need – In library , there is a latent need of handwritten pattern recognizer so that the entries

can be made without making any mistake.

Constant Need- In research area, laboratory, there is constant use of this method to extract the

important data.

Variable Need – There is variable need in Educational Institute like student detail analysis, project

evaluation, exams.

General Need – Postal address interpretation is the general need assigned at postal offices.

Addresses that are handwritten are of poor quality, these are encoded from the image of letters

Niche Need- There is several need in signal processing, image analysis and computer graphic and

Artificial intelligence
 Questionnaire Method

Table 2.2: Questionnaire for the Product

Title
Customer name: SANJAY
AGARWAL

Occupation :Researcher in
laboratary
Address : SAKCHI ,UP
Q. 1 Does this method seem practical to you?
Q. 2 Are you satisfied with the procedure and the techniques used ?
Q. 3 Have you ever worked with this type of technology?
Q. 4 Do you think this method can save your labor?
Q. 5 Do you think this method can save your time?
Q. 6 Do you think there is no loss in this method?
Q. 7 Do you think the method is easy to use?
Q. 8 Is the method fulfilling the need?
Q. 9 Do you need any changes in the mechanism part?
Q.10 How is the accuracy compared to previous methods?
Q.11 Do you think any that its application is limited?
Q.12 Is this method is welfare for the society?
Q.13 Is the method is reliable?
Q.14 Is the maintenance procedure easy and for you?
Q.15 Do you think that is this method is environment friendly?
 Interview Method

Table 2.3: Interview Form for Product

Customer Name:-SANJAY Interviewer:-

AGARWAL SHIVANIKUMARI
Occupation :-RESERCHER AND SHIBANI
Address:-SAKCHI,UP KHATUA
Date:-12/05/20
Questions Customer response to Interpreted Importance
the question need statement rating
by interviewer assigned by
customer
What are the Pattern recgonition by
uses of the machines or computer
current item or which is handwritten
similar items? by human being

Questions that  Accuracy  Highly  Nice

expose the  Reliable accuracy  Good
LIKES of the  Application  Nice
 Handling  More
items or product  Good
 Reduce labour reliable
 Nice
 Vast
Application

 Easy to
handle

 Highly
reduce
labour

Questions  Specfic  Through trial and  Need to

that expose rule for error improved
the structure
DISLIKES of
the product
or item
 Proposition Complexity of model
for rising the structure
incumbent
item or
product

Organizing and Prioritizing of Customer Need

Evaluation shows the best to worst (5-1). Following these weightage , we have to modify
our design.
Table 2.4: Prioritized Customer
Need (Weightage Calculation)

Serial Questions W 5 4 3 2 1 0
(
.No.
5-
1
)
1 Labor and time saving model 4 7 5 6 2 0 0
2 Time saving 3 4 3 6 3 3 1
3 Easy transportation. 4 5 8 6 1 0 0
4 Highly Durable product 3 5 4 3 3 4 1
5 Environment friendly 4 8 6 5 1 0 0
6 Less maintenance cost of the system 4 8 4 5 2 1 0
7 Accuracy 4 8 5 5 2 0 0
8 Easy to handle 3 2 5 3 4 3 3

Interview method for giving weightage to the needs:

 The Customers have given ratings according to their requirements.

 Then the average importance of need is calculated by the following method.

 The weightage scale ranges from 0 to 5.

 5 is the maximum and 1 is the minimum weightage that a customer can rate.

 If customer does not mention any thing then the rating is considered as0.

 Then the average importance of need is calculated in the following manner,

 ( n1 ∗ 5 + n2 ∗ 4 + n3 ∗ 3 + n4 ∗ 2 + n5 ∗ 1 + n6 ∗0)

n1 + n2 + n3 + n4 + n5 +n6

Here n1,n2,n3,n4,n5,n6 = number of customers who gives the corresponding ratings to the product that is

5, 4, 3, 2, 1, 0 respectively. This helps us to identify the needs of the customers which is required to be

given prioritization. In the above method weightage(WT) is given by referring customers about each part

of the product and ask them to rate each function according to their need from 1 to 5 and then the average

importance of need is calculated as follows Example: Weightage calculation of Sl. No. 11is

3∗ 4+4∗ 3+1∗ 2+2∗1

WT=
8

List of Prioritize Need

 Affordable price

 Maintenance free

 Easy to move

 Long lasting
 Chapter 3
Function Decomposition

Work done by each member of Team OF EEE SEC-“C”

Team leader: SHIVANI KUMARI AND SHIBANI KHATUA
Team members Task(s) to be Task assigned Task
performed date complete
d on
SHIVA ANALYSIS AND 29.5.2020 31.5.2020
NI REPORT
KUMA
RI
SHIBANI KHATUA ANALYSIS AND 29.5.2020 31.5.2020
REPORT

TRADITIONAL METHOD

CHARACTER EXACTRACTION
INPUT

CHARACTER RECOGNITION
HANDWRITTEN
PATTERN

FEATURES EXTRACTION

OUTPUT

Fig 1 . Block diagram of traditional method

Page 22
 MODERN METHOD

PRE-PROCESSING FEATURES
EXACTRACTION CLASSIFICATI
ON

SENSOR

OUTPUT

HANDWRITTEN
PATTERN

Fig 2. Block diagram of modern method

CONVOLUTIONAL NEURAL NETWORK

INPUT

CONVULUTIONAL
LAYER

POOLING LAYER

FULLY CONNECTED
LAYER

OUTPUT

Fig 3. Block diagram of convolutional neural network

Page 23
Chapter 4 Product Tear-down

Task Distribution and Monitoring

Work done by each member of Team OF EEE SEC-“C”

Team leader: SHIBANI KHATUA &SHIVANI KUMARI
Team member Task(s) to be Task assigned date Task
performed completed on
1:SHIBANI KHATUA ANALYSIS AND DATA 17.05.2020 19.05.2020
COLLECTION
2:SHIVANI KUMARI ANALYSIS AND 17.05.2020 19.05.2020
DATA COLLECTION

INPUT

CONVOLUTION
AL NEURAL
LAYER

POOLING LAYER

DENSE LAYER

OUTPUT

Fig 4. Block of the whole system

 Chapter 5
Product Architecture

It is the phase wherein we begin to take significant choices on how the Last item will

truly work. It starts the arrangement of powerful frameworks of the following system.

: Engineering Types

There are two kinds of designs: Product and Portfolio. Portfolio plans Relate to a

social event or gathering of things however thing models relate to express thing.

Philosophies for thing setup turn around the thing's business sector and execution.

 Integral:- Essential thing models are physical structures where most of the sub

capacities manual for a single or unobtrusive number of physical parts.

 Modular:- Thing modules are described as essential physical things

foundations that have an organized correspondence with a subset of a thing's

capacity illustrate.

Task Distribution and Monitoring

The assignment for this work has been partitioned among the gathering individuals.

The Team head has allocating various errands to various individuals and set the

timetable for fulfillment of this undertaking as depicted in the table below:-

Work done by each member of EEE-“C”

Team leader: Shibani Khatua
Team member Task(s) Task assigned Task
to be date completed
perform on
ed
1: Shibani Khatua Concept generation 16.05.2020 18.05.2020

2: Shivani Kumari Theorectical 16.05.2020 18.05.2020

evaluation

Page 25
 Task completed with alternative arrangement:
Task name Originally Reason Member,
assigned to for who
Non- complet
compl ed it
etion

Signature of Team Members: Shibani khatua , Shivani Kumari

Member 1:Shibani Khatua Member 2: Shivani Kumari

A convolutional neural network consists of an input and an output layer, as well as

multiple hidden layers. The hidden layers of a CNN typically consist of a series of

convolutional layers that convolve with a multiplication or other dot product. The

activation function is commonly a RELU layer, and is subsequently followed. By

additional convolutions such as pooling layers, fully connected layers and

normalization layers, referred to as hidden layers because their inputs and outputs are

masked by the activation function and final convolution. Though the layers are

colloquially referred to as convolutions, this is only by Convention. Mathematically, it

is technically a sliding dot product or cross Correlation. This has significance for the

indices in the matrix, in that it affects how weight is determined at a specific index

point.

Page 26
CONVOLUTIONAL LAYER

When programming a CNN, the input is a tensor with shape (number of images) x

(image width) x (image height) x (image depth). Then after passing through a

convolutional layer, the image becomes abstracted to a feature map, with shape

(number of images) x (feature map width) x (feature map height) x (feature map

Channels). A convolutional layer within a neural network should have the following

attributes:

 Convolutional kernels defined by a width and height (hyper-parameters).

 The number of input channels and output channels (hyper - parameter).

 The depth of the Convolution filter (the input channels) must be equal to the

number channels (depth) of the input feature map.

POOLING LAYER

Convolutional networks may include local or global pooling layers to streamline the

underlying computation. Pooling layers reduce the dimensions of the data by

combining the outputs of neuron clusters at one layer into a single neuron in the next

layer. Local pooling combines small clusters, typically 2 x 2. Global pooling acts on

all the neurons of the convolutional layer. In addition, pooling may compute a max or

an average. Max pooling uses the maximum value from each of a cluster of neurons at

the prior layer. Average pooling uses the average value from each of a cluster of

neurons at the prior layer.

FULLY CONNECTED LAYER

Fully connected layers connect every neuron in one layer to every neuron in another

Layer. It is in principle the same as the traditional multi-layer perceptron neural

network (MLP). The flattened matrix goes through a full layer to classify the images.
CONSTRUCTING THE ARCHITECTURE OF NETWORK
Neural networks with multiple hidden layers can be useful for solving classification

Problems with complex data, such as images. Each layer can learn features at a

different level of abstraction. However, training neural networks with multiple hidden

layers can be difficult in practice. One way to effectively train a neural network with

multiple layers is by training one Layer at a time. You can achieve this by training a

special type of network also known as an autoencoder for each desired hidden layer.

This experiment focuses on how to train a neural network with two hidden layers to

Classify digits in images. First, you train the hidden layers individually in an

unsupervised fashion using encoding and decoding under the hidden layers. Then you

train a final softmax layer, and join the layers together to form a stacked network,

which you train one final time in a supervised fashion.

DATASET

FIG 6. OVERVIEW OF DIGITS TAKEN UNDER EXPEIMENT

 Chapter 6
Concept Generation

Concept generation enables the team by giving a platform to apply their ideas in order

to generate the solution.

Task Distribution and Monitoring

The task for above work is distributed among various group members. The team leader

has assigning different task to different members of the Group and set the timeline for

this task as shown in table6.1

Table 6.1 Work Distribution

Workdone by each member of the team

Team member Task(s) to be Task assigned Task completed

performed date date
1.Shibani Khatua Memory map, 08.03.2020 15.03.2020
c-sketch, &
concept
variant
2. Shivani Kumari Idea generation 08.03.2020 15.03.2020

Task completed with alternative arrangements:

Task name Originally Reason for Member, who
assigned to non-completion completed it
Signature of team members:
Member 1:Shibani Khatua Member 2:Shivani Kumari
Handwritten pattern recognition played a big role in the technology world then. It also

played an important role in the storage and in the recovery of critical handwriting

Information. This handwritten recognition ensured an accurate medical care and it also

reduced storage costs. It ensured that an essential field of research remains available to

students in the future .Various applications of handwritten pattern recognition were

there like National ID number recognition, postal office automation with code number

recognition on envelope, automatic license plate recognition and bank automation. In

this era of globalization, technologies continue to improve and improve more in no

time.

TRADITIONAL TECHNIQUES

 CHARACTER EXACTRACTION

 CHARACTER RECOGNITION

 FEATURES EXTRACTION

MODERN TECHNIQUES

ONLINE RECOGNITION

 PRE-PROCESSING

 FEATURES EXTRACTION

 CLASSIFICATION
 Chapter 7
Concept Selection

Character recognition from handwritten images has received greater attention in

research area of pattern recognition due to vast applications and ambiguity in the of

learning methods learning methods. Primarily, two steps including character

recognition and feature extraction are required based on some classification algorithm

for handwritten pattern recognition. Previous schemes exhibit lack of high accuracy

and low computational speed for handwritten pattern recognition process. The aim of

the proposed attempt was to make the path toward digitalization should be clearer and

provide high accuracy and faster computational for recognizing the handwritten

pattern.

The present research employed convolutional neural network consisting of different

layers for recognizing (encoding and decoding) and classifying the given input to a

better output. As classifier, MNIST as dataset with suitable parameters for training

and testing and deep learning framework for handwritten pattern recognition. The

convolutional neural network system successfully imparts accuracy up to 99.21%

which is higher than formerly proposed schemes. In addition, the proposed system

reduces computational time significantly for training and testing due to which

algorithm becomes efficient.

CONVOLUTIONAL NEURAL NETWORK

The name “convolutional neural network” indicates that the network employs a

Mathematical operation called convolution. Convolution is a specialized kind of linear

Operation. Convolutional networks are simply neural networks that use convolution in

place of general matrix multiplication in at least one of their layers.

FIG NO-7 OVERVIEW OF CONVOLUTIONAL NEURAL NETWORK

ARCHITECTURE
Chapter 8
Product Embodiment

This stage means to change item ideas and structures into acknowledged Frameworks.

These frameworks must fulfill client needs, be vigorous concerning all natural and

client conditions, and be intended to diminish the probability of disappointment.

Task Distribution and Monitoring

The task for this work has been divided among the group members.

Table 8.1 : Work Distribution

Work done by each member of Team ID 08 of Section F

Team leader:
Task(s) to be Task assigned Task
performed date completed on
1: Shivani kumari Collecting 9.05.2020 26.05.2020
information
2: shibani khatua Collecting 9.05.2020 26.05.2020
information
Project Modelling:-

FIG NO-8 OVERVIEW OF CONVOLUTIONAL NEURAL NETWORK

FRAMEWORK

FIG NO-9 : OVERVIEW OF CONVOLUTIONAL NEURAL

NETWORK
 (a) Convolutional layer (CNL)
Convolutional neural network (CNL) is the first layer in convolutional neural network

which memorizes the features of input image which covers its entire region during

scanning through vertical and horizontal sliding filters. It adds a bias for every region

followed by evaluation of scalar product of both filter values and image regions. For

thresholding element-wise activation function, such as max(0,x), sigmoid and tan(h),

is applied to output of this layer via rectified linear unit.

(a) Pooling layer (PL)

At second, there comes pooling layer which is also called as max pooling layer or sub

Sampling. In pooling layer (PL), shrinkage in the volume of data takes place for the

easier and faster network computation. Max pooling and average pooling are main

tools for implementing pooling .This layer obtains maximum value or average value

for each region of the input data by applying vertical and horizontal sliding filters

through input image and reduces the volume of data .

(b) Fully connected layer or dense layer

At last, there is fully connected layer after convolution and pooling layer in the

standard neural network (separate neuron for each pixel) which is comprised of n

numbers of neurons, where n is the predicted class number.

For example, There are ten neurons for ten classes (0–9) in digit character

classification problem. However, there should be 26 neurons for 26 classes (a–z)

for English character Classification Problem However, deep neural network

architecture consists of many nonlinear hidden layers with a enormous number of

connections and parameters. Therefore, to train the network with very less amount

of samples is a very difficult task .In convolutional neural network ,

Page 35
 only few set of parameters are needed for training of the system. So, convolutional

neural network is the key solution capable map correctly datasets for both input

and output by varying the trainable parameters and numbers of hidden layers with

high accuracy. Hence, in this work, convolutional neural network architecture

with deep learning framework had been considered as the best fit for the character

recognition from the handwritten pattern images. For the experiments and the

verification of the system’s performance, the normalized standard MNIST dataset

is utilized.

FIG NO-10 Architecture of the convolutional neural network

FIG NO-11 Control flow diagram

Page 36
 Chapter 9
Analytical and Numerical Model
Solutions

The idea is hypothetically broken down, with the goal that numerous setups might be

looked to acknowledge favored arrangements. Various computational arrangement

strategies exist to fathom item models.

Task Distribution and Monitoring

The errand for this work has been separated among the gathering individuals. The Team

head has doling out various assignments to various individuals and set the timetable for

fruition of this undertaking as appeared in Table 11.1 :

Table 9.1 : Work Distribution

Work done by each member of EEE –“C”

Team leader:
Team member Task(s) to be Task assigned Task
performed date completed on
1: Shivani kumari Theoretical 10.04.2020 17.04.2020
Analysis
2:Shibani khatua Theoretical 10.04.2020 17.04.2020
Analysis

THEORETICAL ANALYSIS

FUNCTION AND PARAMETERS

The stacked neural network is a simple three-layer neural network including an

encoding layer and a decoding layer for the system function where output units are

directly connected back to input units that shown in Figure 1. The proposed sparse

neural network was trained on the 𝑋𝑛𝑙 raw inputs , 𝑋𝑚𝑙 hidden layer and output layer

Page 37
𝑌𝑛𝑙 where n is number of inputs or outputs neuron and m is number of hidden neuron

and l is number of sparse neural network . The output layer maps the input vector 𝐼𝑛𝑙

to the hidden layer 𝐻𝑚𝑙with a non-linear function S:

------------------------- (1)

Where, 𝑊𝑖𝑙 denote the parameters (or weights) associated with the connection

between input unit and hidden unit bm are a biases in hidden layer. S(v) is the sigmoid

function. The sigmoid function is defined as S(v) consisting of inverse exponential

formula.

The output layer 𝑌𝑛𝑙 has the same number of units with the input layer and defined by

a power series expansion equation. Where 𝑊j, denote the parameters (or weights)

associated with the connection between hidden unit and output unit.bn are the biases in

the output layer. S is a sigmoid function in the exponential form.

FIG NO-12 OVERVIEW OF ENCODING AND DECODING LAYER

Page 38
STACKED NETWORK

We introduce the design of digit-level stacked layered network for digits classification.

The first sparse network structure contains the input layer 𝑋𝑛𝑙 to learn primary

features on the raw input that illustrated in below figure. The first sparse structure

produces the primary feature (I). The primary feature𝐻𝑚1 feeds the input layer into

the second trained sparse network that produce the secondary features (II). In below

figures focuses the primary features used as the raw input to next sparse network to

learn secondary features. Then, the secondary feature treat as input layer to a softmax

classifier to map secondary features to digit labels that shown in below figures.

FIG NO-13 OVERVIEW OF HIDDEN LAYERS SHOWING PRIMARY AND

SECONDARY FEATURE

FIG NO-14 OVERVIEW OF SOFTMAX CALSSIFIER LAYER

Page 39
OVERALL STRUCTURE

FIG NO-15 OVERVIEW OF PROPOSED NETWORK STRUCTURE

FIG NO-16 OVERVIEW OF CONVOLUTIONAL NETWORK STRUCTURE

Page 40
TRAINING

FIRST SPARSE NETWORK

The training begins by a sparse neural network on the training data without using the

labels. An autoencoder is a neural network which attempts to replicate its input at its

output. Thus, the size of its input will be the same as the size of its output. Neural

networks have weights randomly initialized before training. Therefore the results from

training are different each time.

FIG NO-18 OVERVIEW OF FIRST SPARSE NETRWORK LAYER STRUCTURE

SECOND SPARSE NETWORK

After training the first sparse network, you train the second network in a similar way.

The main difference is that you use the features that were generated from the first

network as the training data in the second sparse network. Also, you decrease the size

from the first one.

FIG NO-19 OVERVIEW OF SECOND SPARSE NETRWORK LAYER

STRUCTURE

Page 41
SOFTMAX LAYER

Train a softmax layer to classify the 50-dimensional feature vectors. Unlike the sparse

network, you train the softmax layer in a supervised fashion using labels for the

training data.

FIG NO-20 OVERVIEW OF SOFTMAX LAYER STRUCTURE

FORMATION OF NEURAL NETWORK

The formation of neural network takes place by the combination of all network layers

along with softmax layer for the possible outcome

FIG NO-21 OVERVIEW OF DIFFERENT NETWORK LAYER STRUCTURE

Page 42
FIG NO-22 OVERVIEW OF COMBINATIONAL STRUCTURE

With the full network formed, you can compute the results on the test set. To use this,

you have to reshape these into a matrix. It is done by stacking the columns of an image

to form a vector, and then forming a matrix from these vectors.

Page 43
Chapter 10 Bill of materials

Sl. Components Components Material Total Cost

No. to buy from have to be Cost
Market Manufactured
in the workshop
1 Image processing - - Rs 0/-

2 Image - - Rs 0/-
Segmentation
3 Feature Extraction - - Rs 0/-

4 Trained Neural - - Rs 0/-

Network
6 Recognized - - Rs 0/-
Character

Page 44
 Chapter 11 Results

These are the required outputs by the given MATLAB code proposed by us and

visualized by a confusion matrix :-

Page 45
Page 46
 Chapter 12
Conclusion

The aim of our project is to make an interface that can be used to recognize human

handwritten characters .We approached to our problem using convolutional neural

networks, so that we get a higher accuracy than the primitive method. Using modern

day techniques like neural networks to implement deep learning to solve basic tasks

which can be done within a blink of an eye by any human like text recognition is just

scratch the surface of the potential behind machine learning. There are vast

possibilities and use of this technology. Traditional optical character recognizer used

to work similar to biometric device .Photo sensor knowledge was used to gather the

match points of physical attributes and then convert it into database of identified types.

But with the help of recent techniques like convolution neural networks we are

capable of scanning and understand words with an accuracy never seen before in the

past. Several application of handwritten pattern recognition using convolutional neural

network are there like reading postal addresses, bank check amounts and forms.

INDIVIDUAL AND GROUP LEARNING :-

Through this project I learned as an individual team member that

(a) I should vigorously took part in meetings and share information with ideas.
(b) I must be enthusiastic.
(c) I should carefully work.
(d) I should respect other’s contribution.
(e) I must be dedicated to team goal.

(f) I should carry out the tasks through observing process and information.

REFERENCES:-

Page 47
 Surya Nath R S M, Afseena “Handwritten Recognition- A review”,
International Journal of Scientific and Research Publication

 Anita Pal and Davashankar Singnh,”Handwritten English Character

Recognition Using Neural Network”, International Journal of Computer
Science and Communication.

 Manoj Sonkusare and Narendra Sahu “A survey on handwritten character recognition

(hcr) techniques for english alphabets” Advances in Vision Computing: An
International Journal (AVC)

Page 48
Page 49