# Artificial Neural Network Based Optical Character Recognition

**Authors:** Vivek Shrivastava, Navdeep Sharma  
**arXiv:** [1211.4385](https://arxiv.org/abs/1211.4385)  
**Date:** 2012-11-19  
**Summary page:** https://www.summarizepaper.com/en/paper/1211.4385v1/

## Key points

- OCR focuses on the recognition and classification of characters from an image
- Topological and geometrical properties are calculated to improve accuracy
- Human psychology plays a role in perceiving characters based on shape and features
- Spatial pixel-based calculations extract features from images
- Extracted features are organized into vectors
- Artificial Neural Networks (ANNs) are used for character classification in OCR
- ANNs learn to associate feature vectors with corresponding characters through training with labeled datasets
- The study "Artificial Neural Network Based Optical Character Recognition" explores the use of ANNs in OCR technology
- Feature vectors derived from spatial pixel-based calculations can be utilized by an ANN for accurate character recognition

## Summary

The field of Optical Character Recognition (OCR) focuses on the recognition and classification of characters from an image. In order for OCR to be accurate, various topological and geometrical properties are calculated, which are then used to classify and recognize each character. Additionally, human psychology plays a role in perceiving characters based on their overall shape and features such as strokes, curves, protrusions, and enclosures. To extract these properties, also known as features, from an image, spatial pixel-based calculations are employed. These calculations help identify specific characteristics that define each character uniquely. These extracted features are then organized into collections called vectors. One powerful tool used in OCR is an Artificial Neural Network (ANN). ANNs utilize these feature vectors to accurately classify and recognize characters. By training the ANN with a large dataset of labeled characters, it can learn to associate specific feature vectors with corresponding characters. In a recent study titled "Artificial Neural Network Based Optical Character Recognition," authors Vivek Shrivastava and Navdeep Sharma delve into the intricacies of OCR. They highlight the importance of calculating topological and geometrical properties for accurate recognition while considering how humans perceive characters based on their overall shape and features. This research contributes to advancements in OCR technology by exploring the use of artificial neural networks in character classification. The authors provide insights into how feature vectors derived from spatial pixel-based calculations can be utilized by an ANN to accurately recognize characters. Overall, this study sheds light on the complex process involved in optical character recognition and offers valuable insights into improving its accuracy through the use of artificial neural networks.