Scribd
Upload
2 views

HANDWRITTEN_NUMBER_GUESSING

The document outlines a project focused on developing an automated system for recognizing handwritten digits using a feedforward backpropagation neural network. It highlights the challenges of traditional recognition methods and presents objectives aimed at achieving over 95% accuracy. The study concludes that neural networks are effective for this task, with potential for future improvements through deep learning models.

Uploaded by

dinesh.23mic7104
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
2 views

HANDWRITTEN_NUMBER_GUESSING

The document outlines a project focused on developing an automated system for recognizing handwritten digits using a feedforward backpropagation neural network. It highlights the challenges of traditional recognition methods and presents objectives aimed at achieving over 95% accuracy. The study concludes that neural networks are effective for this task, with potential for future improvements through deep learning models.

Uploaded by

dinesh.23mic7104
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 16

.

.
Handwritten number guessing USING
NEURAL NETWORK
• Faculty : Dr. AMGOTHU SHANTHI
• Team Members
• Dinesh Kumar-23MIC7104
• Ramparinitha-23MIC7103
• Esha Madhuri Chekuri - 23BCE7580
• Supriya kowda 23BCE9661
OVERVIEW
-TITLE
INTRODUCTION
PROBLEMSTATEM
ENT OBJECTIVE
LITERATURE
SURVEY BLOCK
DIAGRAM
BASE PAPERS
RESEARCH
PAPERS
REFERENCE
CONCLUSION
1. Introduction
• In recent years, digital recognition has gained importance in various domains
such as banking, security, and postal services. With advancements in artificial
intelligence (AI) and deep learning, neural networks have proven to be an
effective solution for recognizing handwritten characters and digits.
Traditional methods like template matching, decision trees, and statistical
analysis have limitations in handling variations in handwriting styles. The
primary objective of this study is to develop an automated system capable of
recognizing isolated handwritten digits. The system will process digital
images, enhance their quality, and classify the extracted digits using a neural
network. The proposed method employs a *feedforward backpropagation
neural network*, trained with a dataset of handwritten Arabic digits.
Key Applications:

- Banking: Recognizing handwritten


amounts in checks for automatic
processing
- License Plate Recognition: Identifying
vehicle registration numbers
- Postal Services: Automating mail
sorting based on handwritten zip
codes
- Academic Fields: Developing AI-based
examination scoring systems
LITERATURE SURVEY
• 1. Author: Jurgen Frank (1997) -
• Model Used:Polynomial Classifier Approach
• Methodology: - Utilized three classifiers: simple linear, polynomial, and feature-
extracted classifiers
• Karhunen-Loève transformation was applied for feature extraction - Iterative
learning techniques improved accuracy
• 2. Author: Alceu de Britto et al. (2003)
• Model Used: Two-stage Hidden Markov Model (HMM)
• Methodology: - The first stage involved implicit segmentation with multiple
hypotheses - The second stage re-ranked and verified the segmented digits
3. Author: Kussul & Baidyk (2006)
Model Used: Limited Receptive Area (LIRA) Classifier
Methodology: - Trained a three-layer neural network -
Used datasets like MNIST for handwritten digit recognition
- Applied random local descriptors to improve accuracy

4. Author: Westall (1997) -


Model Used : Neural Networks for Digit Segmentation
Methodology: - Proposed a neural network-based
segmentation approach - Implemented in a commercial
system for bank check recognition
Problem Statement:

• Handwritten digit recognition presents a


challenge due to variations in writing
styles, orientation, and image quality.
Traditional recognition techniques often fail
in accuracy, requiring a more robust AI-
based solution.
Objectives:
• 1. Develop an automated system for recognizing
handwritten digits.
• 2. Preprocess input images to enhance quality and
remove noise.
• 3. Implement a neural network model with feedforward
backpropagation for classification.
• 4. Achieve high accuracy (above 95%) on a test dataset.
• 5. Optimize model performance for real-time applications.
ARCHITECTURE
BLOCK DIAGRAM
BASE PAPERS
• https://arxiv.org/pdf/2106.12614
• https://ieeexplore.ieee.org/document/9720854
• https://ieeexplore.ieee.org/document/8109042
• https://ieeexplore.ieee.org/document/9213619
• https://link.springer.com/chapter/10.1007/978-981-19-314
8-2_67
RESEARCH PAPERS
• https://www.academia.edu/download/67807777/jcssp.2009
.427.pdf
• http://jice.um.edu.my/index.php/MJCS/article/view/6216
• https://www.academia.edu/download/67807777/jcssp.2009
.427.pdf
• https://www.worldscientific.com/doi/abs/10.1142/S021800
1497000317
•
https://www.sciencedirect.com/science/article/pii/S089360
8011003170
References
• 1. Saleh Ali K. Al-Omari, Putra Sumari, Sadik A. Al-Taweel, and
Anas J.A. Husain, Digital Recognition Using Neural Network,
Journal of Computer Science, 2009.
• 2. Jurgen Frank, Handwritten Digit Recognition Using
Polynomial Classifier Approach, 1997.
• 3. Kussul, E. & Baidyk, T., Neural Classifier LIRA for
Handwritten Recognition, 2006.
• 4. Westall, J.M., Neural Networks for Handwritten Digit
Segmentation, 1997.
Conclusion

• This study demonstrates the effectiveness


of neural networks in handwritten digit
recognition. The feedforward
backpropagation algorithm successfully
classified isolated digits with a 95%
accuracy rate. Future enhancements could
include deep learning models (e.g., CNNs)
for even higher accuracy and robustness
THANK YOU

You might also like