Final Seminar Presentation2

KJEI’S
TRINITY ACADEMY OF ENGINEERING,PUNE
Seminar Presentation
Department of Information
Technology
1
SPPU
Savitribai Phule Pune
University
Year 2024-2025
Seminar Presentation
“Handwritten Character
Recognition”
2
Presented By Under The
Shruti Mahadik Guidance Of
Mrs. S.A.Joshi
3 CONTENTS
 Problem statement
 Introduction
 Aim and Objective
 Scope
 Literature review
 Architectural Block Diagram
 Mathematical Model
 DFD/ ER Diagram
 UML Diagram
 Algorithms (Compulsory include)

4  Conclusion
 Future Scope
 References
5  Problem statement
Develop a handwritten character recognition system that utilizes neural networks, leveraging
diagonal-based feature extraction techniques to enhance accuracy. The system should analyze
and classify handwritten English characters from images.
 Introduction
Handwritten character recognition has gained significant attention due to its applications in
various fields, including document digitization and automated data entry. This study explores
advanced techniques, particularly neural networks, to enhance recognition accuracy. Drawing
insights from research on diagonal-based feature extraction, the system aims to capture
distinctive characteristics of handwritten characters.
6  Aim
The aim of handwritten character recognition (HCR) is to convert handwritten text into
machine-readable text to reduce manual data entry.
 Objective
To Achieve high accuracy in recognizing and interpreting handwritten characters and words
despite variations in handwriting styles and quality.
 Scope
The scope of handwritten character recognition (HCR) includes automating the conversion of
handwritten text into digital formats for efficient data management and enhancing
accessibility across diverse applications
 Literature review
7
Paper Publish year Authors

1. A Survey of 4 August 2019 Alejandro Baldominos Yago Saez
Handwritten Character Pedro Isasi
Recognition with MNIST
and EMNIST
2. Handwritten English 2010 A Pal, D Singh
Character Recognition
Using Neural Network
3. Diagonal based feature 2011 3rd International Conference J. Pradeep; E. Srinivasan; S.

extraction for handwritten on Electronics Computer Himavathi
character recognition Technology
system using neural
network
 Architectural Block Diagram
8
Handwritten Preprocessing Feature Vector

Image(Input Data) Binarization
Extraction Representation
Evaluation & Neutral Model

Post Processing Recognition Output
Training
Accuracy
 Mathematical Model
9
The mathematical model for handwritten character recognition typically uses a neural
network, often a Convolutional Neural Network (CNN). Each input image is represented as a
vector x/{x}x of pixel values.
During the forward pass, the output of each layer is computed as:
h(k)=f(W(k)⋅h(k−1)+b(k))
where f is an activation function, W is the weight matrix, and b is the bias vector. The output
layer uses the softmax function to generate probabilities for each character class.
The model's performance is evaluated using the cross-entropy loss function:
L=−i=1∑Cyilog(y^i)
Backpropagation updates the weights to minimize this loss.
 Algorithms
10
1] Convolutional Neural Networks (CNNs) :
A CNN is a type of deep learning model designed to process images. It learns to identify
patterns and features in images, making it great for recognizing handwritten characters.
•Architecture: Consists of layers like convolutional, pooling, and fully connected

layers.
•Convolutional Layers: Apply filters to extract features (edges, textures) from images.
•Pooling Layers: Downsample the feature maps to reduce dimensionality while
retaining important information.
•Activation Functions: Typically use ReLU to introduce non-linearity.
•Training: Involves backpropagation and optimizers like Adam to adjust weights based
on loss
11
2] Support Vector Machine

SVM aims to find the optimal hyperplane that separates different classes in the feature space.
The hyperplane is selected to maximize the margin between the closest data points of each
class, known as support vectors
• Classification Method: SVM is used for binary and multi-class classification tasks.
•Optimal Hyperplane: Finds the best boundary (hyperplane) to separate different classes in
high-dimensional space.
•Support Vectors: Focuses on data points closest to the hyperplane, known as support
vectors, which influence the model.
•Kernel Trick: Utilizes kernel functions (e.g., linear, polynomial, RBF) to handle non-
linearly separable data by transforming it into higher dimensions.
12  Conclusion
Handwritten character recognition has made significant strides through advanced techniques
like neural networks and innovative feature extraction methods. Research shows that
leveraging datasets such as MNIST and EMNIST can greatly enhance recognition accuracy.
Continued exploration in this field promises further improvements, making systems more
reliable and efficient for real-world applications.
 Future Scope
The future of human character recognition (HCR) lies in developing more sophisticated
algorithms that can adapt to diverse handwriting styles and languages. Enhanced real-time
processing capabilities will enable seamless integration into mobile applications. There is
potential for incorporating artificial intelligence to improve accuracy and reduce biases.
Furthermore, integrating HCR with augmented reality could create innovative user
experiences.
 References
13
[1] R. Plamondon and S. N. Srihari, “On-line and off- line handwritten character recognition:
A comprehensive survey,”IEEE. Transactions on Pattern Analysis and Machine Intelligence,
vol. 22, no. 1, pp. 63-84, 2000.
[2] S. Impedovo, L. Ottaviano and S. Occhinegro, “Optical character recognition”,

International Journal Pattern Recognition and Artificial Intelligence, Vol. 5(1-2), pp. 1-24,
1991
[3]Calderón, A.; Roa-Valle, S.; Victorino, J. Handwritten digit recognition using convolutional
neural networks and Gabor filters. In Proceedings of the 2003 International Conference on
Computational Intelligence, Cancun, Mexico, 19–21 May 2003.
14
Thank You

Final Seminar Presentation2

Uploaded by

Copyright:

Available Formats

Final Seminar Presentation2

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Final Seminar Presentation2

Uploaded by

Copyright:

Available Formats

KJEI’S

TRINITY ACADEMY OF ENGINEERING,PUNE

 Aim and Objective

 Architectural Block Diagram

 Algorithms (Compulsory include)

Paper Publish year Authors

3. Diagonal based feature 2011 3rd International Conference J. Pradeep; E. Srinivasan; S.

Handwritten Preprocessing Feature Vector

Evaluation & Neutral Model

•Architecture: Consists of layers like convolutional, pooling, and fully connected

2] Support Vector Machine

[2] S. Impedovo, L. Ottaviano and S. Occhinegro, “Optical character recognition”,

You might also like