Neural network based handwritten hindi character recognition system

Abstract

Neural Networks are being used for character recognition from last many years but most of the works were reported to English character recognition. Character recognition is one of the applications of pattern recognition, which has enormous scientific and practical interest. Many scientific efforts have been dedicated to pattern recognition problems and much attention has been paid to develop recognition system that must be able to recognize a character. The main driving force behind neural network research is the desire to create a machine that works similar to the manner our own brain works. Neural networks have been used in a variety of different areas to solve a wide range of problems. A very little work has been reported for Handwritten Hindi Character recognition. In this paper, we have implemented Gradient feature extraction technique, which provides more than 94% recognition accuracy. We have acquired 1000 samples of handwritten Hindi characters by initializing the mouse in graphics mode. The 500 samples have been used for training the network (Train Data) and remaining 500 samples have been used for testing the network (Test Data). The system has been trained using several different forms of handwritings provided by both male and female participants of different age groups. Finally, this rigorous training results an automatic HCR system using MLP network. The error backpropagation algorithm has been used to train the MLP network. A comparative analysis was performed by implementing both global input and Gradient feature input. We have concluded that gradient feature extraction technique provides better recognition accuracy with reduced training time.

Hindi character recognition using RBF neural network and directional group feature extraction technique

“…This language consists of 49 characters and is written from left to right. A set of handwritten Hindi character is given in following Figure1 [2,5,7,12,14]. [2,5,7,12,14] …”

“…The collected hard copy samples have been scanned through scanner and stored in computer system for training and testing purpose. For normalization, different samples of different size are converted in a 30x30 pixel box [2,5,7] III. GRADIENT FEATURE EXTRACTION For handwritten character recognition with better recognition accuracy, a robust feature extraction technique has become very important.…”

“…Gradient feature extraction technique is one of them. [1,5,7,12,14] A. Gradient Feature Extraction In gradient feature extraction, gradients are computed with the help of Sobel operator. There are two templates in Sobel operator named as horizontal and vertical template.…”

“…There are two templates in Sobel operator named as horizontal and vertical template. The templates are given below in Figure2 [5,7,12].…”

“…This 30x30 size of character is padded by a boundary with zeros to make it 32x32 sizes. The two gradient components at location (i, j) are calculated by the following equations [5,7,12] …”