Volume 7, Issue 7, July – 2022 International Journal of Innovative Science and Research Technology

ISSN No:-2456-2165

Personality Analysis through Graphology

using Image Processing
Dr. Deepu R., Suman K.M., S. S. Surabhi, Nischal S., Nisarga P.
Department of Computer Science and Engineering,
ATME College of Engineering, Mysuru

Abstract:-Graphology is a method of identifying, (CNN) is a popular technique used for image recognition.
evaluating and understanding human personality traits in Dense CNN is used to extract the features from handwritten
person by analysing the strokes and patterns revealed by text sample.In order to enhance the accuracy of analyzing the
their handwriting. Handwriting shows a person’s genuine handwriting image.
nature, including emotional outlay, fears, honesty,
defences and many other traits. Graphologists, who A. Motivation
analyse handwriting professionally, usually determine the Analyzing someone's handwriting is one way among
writer from a piece of handwriting. The analyst's level of several method to understand the psychology of a person.
expertise affects how accurate the analysis is. Despite Some applications of graphology are listed here.
being effective, human assistance in handwriting analysis  Psychological analysis: Psychologists and counsellors
is expensive and error-prone. Hence, the proposed employ graphology in clinical testing. However, it is
methodology focuses on developing a system that employs typically used in conjunction with other projective
machine learning to predict personality traits without the personality assessment tools rather than on its own.
need for human interaction. To make this happen, we  Employment profiling: For recruitment purposes,
look at handwriting features in a document to predict a companies utilize handwriting analysis. A graphological
writer's personality traits.After extracting all required report should be used in conjunction with additional
features from the image containing the handwriting, resources such as inbackground checks, practical
DenseNet are trained which output personality trait of demonstration or record of work skills.
the writer. The emotions are depression, anxiety, panic,
stress and no sign of psychological issues are been B. Problem Statement
predicted. For the project, 4899 images of handwriting Personality analysis through graphology using image
samples have been acquired. processing. The main aim of this project is to examine and
evaluate a person’s personality and character by an analysis
Keywords:-Graphology, Human Personality, Psychological of their handwriting.
Analysis, Image Processing, Machine Learning, Densely
Connected Convolutional Network. C. Existing System
Earlier, before the recent advances in deep learning and
I. INTRODUCTION image processing,graphologists had to manually analyze
handwriting in order to identify a person’s traits.
Graphology is defined as the analysis of the physical  Based on the handwriting, graphologists used to examine
characteristics and patterns of the handwriting of an the features as forensic documents.
individual to understand his or her psychological state at the  Now there are features that allow us to understand emotion
time of writing. Handwriting is a kind of projective test of the person from their handwriting simply by scanning it
where the unconscious comes to the fore and expresses itself or capturing it with cameras.
in the conscious. It works on the principle that while writing  Since there may be some errors and inaccuracies when
our hand is controlled by our subconscious mind. A writer evaluating the handwriting, this process requires an
does not consciously draw each letter by his or her hand extensive number of human resources.
while writing, just like how a program does not consciously
remember and locate the position of each letter on a Thus, manual examinations are often performed using a
computer keyboard while typing. These graphic movements limited sample of data.
generated by the subconscious mind reflects the state of the
subconscious itself. Humans have always been intrigued by D. Proposed System
variability and uniqueness of each individual. A Graphologist A system is proposed that the basic handwriting analysis
roughly interpret an individual’s character and personality tasks of graphology to be determined in order to identify a
traits by analyzing the handwriting. We can use graphology few important psychological emotions. From a sample
to determine the personality and character profile of a person. handwriting image, five traits or emotions are considered to
beextracted from it. Each dense convolutional network will
Handwriting analysis helps in identifying these be trained for a particular personality trait using an
emotions through machine learning. Approaches such as integration of these discrete data. The classifier will then be
random forests, neural networks, Bayesian networks, k-NN able to predict the emotion of the writer.
and SVM are implemented for recognition of emotional state.
Emotions can also be predicted through psychological
analysis and graphology.ConvolutionalNeural Network

IJISRT22JUL874 www.ijisrt.com 211

Volume 7, Issue 7, July – 2022 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
II. METHODOLOGY  A filter (fh×fw×d)
 Results in a dimension (h-fh+1)× (w-fw+1) ×1
A. System Description
The block diagram of the proposed system is shown in C. The Non-Linear Layer
figure 1. The writer gives the handwriting sample in the form It is adder after each convolution process. Without this
of image. The emotions database is built in accordance with trait, a network would not be sufficiently intense and unable
the DASS test results. The database's numerous classes are to simulate the response variable. It has an activation
involved in this. The scanned image is then preprocessed function that offers a non-linear property.
using image resizing and normalization. The training dataset
and validation dataset are used to train and validate the D. The Pooling Layer
convolutional neural network algorithm.Thetesting dataset is Similar to the non-linear layer, it also advances in that
then used to test it. The CNN model then receives a fresh direction. It uses the width and height of the image and
input image and, using features from the image, the performs a down sampling technique to them. The size of the
model predicts the emotional state of the image.Todetermine image is reduced as a consequence. This implies that a
the accuracy, the predicted emotion is compared to the results detailed image is reduced into smaller images if some
of the DASS analysis. Therefore, if necessary, psychological characteristics were previously detected during the previous
counselling or medication is advised. convolution operation and the detailed image is no longer
needed for further processing.
Data Input
E. Fully Connected Layer
After the succession of convolution, non-linear, and
Preprocessing pooling layers is finished, it is crucial to link an overall
linked layer. The output data from the convolution network is
Feature Extraction sent into this layer. A fully connected layer creates an N-
dimensional vector when it is coupled to the network end,
Training the Classifier where N is the number of classes from which the model
selects the required class.

Predict Personality Traits F. CNN Model

 This CNN model was developed using the TensorFlow
framework and the OpenCV library and is widely
Fig. 1: An overview flow chart for the proposed system utilized in real-time applications.
The CNN algorithm is employed and cross verified to  This approach may also be applied to the development
achieve the same result. Using complementary subsets of the of comprehensive software that monitors everyone
dataset for evaluation and training, the cross-validation entering at public gatherings.
approach evaluates machine learning models. K-fold cross-
G. Layers in CNN Model
validation splits the dataset into k equal parts for this CNN
Convo2D Layer: It uses the ‘ReLu’ activation algorithm
model, one of which is maintained for validation and the
and contains 100 filters. Rectified Linear Unit, or ReLu, is a
other portions of which are utilized for training.
function that outputs directly if the input is positive and zero
B. Convolutional Layer otherwise.
This layer is always deployed first. It receives the image
MaxPooling2D: It is utilized with 2×2 pool or filtration
(a matrix of pixel values). Assume that the top left corner of
sizes.
the image is where the input matrix's response begins. The
software then selects the smaller matrix there, known as a Flatten() Layer: All the layers are flattened into a single
filter.Convolution is then generated by the filter and is 1D layer using it.
applied to the input image. Multiplying the original pixel
values by the filter's value is its task. A single number is Dropout Layer: It functions as a barrier against the
produced when all of these multiplications are combined model overfitting.
together. Because it only reads the upper left corner of the
picture, the filter moves.A unit on the right also performs a Dense Layer: SoftMax is the activation function used
similar task. After the filter has processed each point, a here, and it will output a vector with two values from the
matrix is produced, but it is less than the input matrix.Froma probability distribution.
human perspective, this procedure is comparable to
recognizing basic colours and visual borders. However, the DenseNet-201 is a 201-layer convolutional neural
full network is necessary in order to identify the fish. The network. It is shown in Figure 2. The ImageNet database
network will include a number of convolutional layers with contains a pretrained version of the network that has been
nonlinear and pooling layers.Convolutionis the initial layer to trained on more than a million images.
extract features from an input image. Convolution uses tiny
squares of input data as its input. An image matrix and a filter
or kernel are the two inputs for the mathematical process.
 Dimension of an image matrix (h×w×d)

IJISRT22JUL874 www.ijisrt.com 212

Volume 7, Issue 7, July – 2022 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
anxiety, stress and excludes no sign of psychological
issues using handwritten words.

IV. FUTURE WORKS

The project has potential for further enhancement. Here

Fig. 2: Schematic diagram of DenseNet-201 model is a list of some enhancements that may be made:
 More data can help machine learning algorithms be trained,
therefore gather more handwriting samples from more
writers.
 Improve the algorithms to handle difficult handwriting
scenarios for greater accuracy.
 Develop more algorithms to extract additional
characteristics so that more personality traits may be
identified.
Fig. 3: Architecture of Densely Connected Convolutional  Other machine learning algorithms such as KNN, Decision
Network Tree, ANN, etc. are trained and compared.
III. RESULT
 Develop a graphical user interface to make the task simple
for anybody to utilize.
A crucial component of every neural network model is
the loss function. It demonstrates how projected value and V. CONCLUSION
actual value diverge. There will be loss and validation Our project proposes an automated method to predict a
accuracy for each fold for the cross-validation models on the person’s personality from a sample of their handwriting using
validation dataset. The average degree of accuracy was CNN algorithm.Theproposed system may be utilized by the
95.27%. The accuracy for the training dataset and validation graphologist as an additional tool to improve the accuracy of
dataset as determined by the cross-validation approach is handwriting analysis and also make the process fast. A
shown in Figure 4.Thenumber of epochs and loss/accuracy person's handwriting may be taken into consideration as a
are displayed on the X and Y axes, respectively. With significant indicator of their mental health. The same goal is
increasing epoch values, the colored lines, or validation accomplished using machine learning. In order to extract
accuracy, increase higher. characteristics from handwriting samples and classify them
into the relevant groups, convolutional neural network
technique is utilized. The number of training samples rises
when cross-validation is utilized, improving accuracy.
During training, the cross-validation models' best weights are
stored.

REFERENCES

[1.] Handwriting Analysis Through Graphology: A Review,

Lakshmi Durga, Deepu R – International Conference on
Advances in Computing, Communications and
Informatics (ICACCI), (2018).
[2.] Personality analysis through handwriting recognition,
Ajeet Ram Pathak, Abhishek Raut, Soham Pawar,
Mansi Nangare, Harshdeep Singh Abbott & Pawan
Chandak – Journal of Discrete Mathematical Sciences
Fig. 4: Loss and Accuracy for the training dataset and Cryptography (2020).
[3.] Handwriting based personality analysis: an image
By using machine learning to analyse a person's processingbased tool, Pratik Oak, Subodh
handwriting patterns, a method has been proposed to predict Pandharkame, Sandeep Munghate, VaradaKolge –
some psychological emotions and personality characteristics. International Conference on IoT Based Control
Five handwriting characteristics could be extracted, and any Networks and Intelligent Systems (2021).
one among those five personality traits could be predicted [4.] Recognition of Emotional State Based on Handwriting
using various combinations of those characteristics. Each Analysis and Psychological Assessment, S. V. Kedar,
neural network classifier is trained for each of the personality Shilpa Rokade – International Journal of Engineering
traits. After sufficient of training, we can predict personality and Advanced Technology (IJEAT), (2019).
characteristics on fresh samples of handwriting quite [5.] DenseNet-201-Based Deep Neural Network with
accurately and effectively.Thisdemonstrates that using an Composite Learning Factor and Precomputation for
ensemble of the best cross-validation model weights Multiple Sclerosis Classification, Shuihua Wang, Yu-
increases classifier accuracy. Finally, we draw the conclusion Dong Zhang – ACM Transactions on Multimedia
that we can predict negative emotions like depression, panic, Computing, Communications and Applications (2020).

IJISRT22JUL874 www.ijisrt.com 213

Volume 7, Issue 7, July – 2022 International Journal of Innovative Science and Research Technology
ISSN No:-2456-2165
[6.] Machine learning approach for personality analysis and
writer identification through handwriting, Vishal Patil,
Harsh Mathur – International Conference on Inventive
Computation Technologies (ICICT), (2020).
[7.] Personality Type Assessment System by using
Enneagram-Graphology Techniques on Digital
Handwriting, D. Pratiwi, G. Santoso, and F. Saputri –
International Journal of Computer Application (0975-
8887), 2016, Vol 147, No.11.
[8.] Survey on handwriting-based personality trait
identification, K. Chaudhari and A. Thakkar – Expert
Syst. Appl., vol. 124, pp. 282–308, 2019.
[9.] Is Graphology Useful in Assessing Major Depression?
M. Giannini, P. Pellegrini, A. Gori, and Y. Loscalzo –
Psychol. Rep., vol. 122, no. 2, pp. 398–410, 2019.
[10.] Handwriting based Personality Identification using
Textural Features, AbdellatifGahmousse,
AbdeljalilGattal, ChawkiDjeddi and Imran Siddiqi –
International Conference on Data Analytics for
Business and Industry: Way Towards a Sustainable
Economy (ICDABI) (2020).

IJISRT22JUL874 www.ijisrt.com 214