Recognition Offline Handwritten Hindi Digits Using Multilayer Perceptron Neural Networks

Handwritten Hindi digit recognition plays an important role in eastern Arab countries especially in the courtesy amounts of Arab bank checks. In this paper, we proposed an efficient offline handwritten Hindi digits recognition system and developed using Multilayer Perceptron Neural Network (MLP). The implemented system recognizes separated handwritten Hindi digits scanned using a scanner. The system has been designed, implemented and tested successfully. The error backpropagation algorithm has been used to train the MLP network. An analysis shows increasing in the recognition rate by doing some enhancements. The proposed system has been trained on samples of 1000 images and tested on samples of 600 images written by different users selected from different ages. The samples were scanned and preprocessed before entering the Neural Network for recognition. The An experimental result showed a very good recognition rate in comparison with other Hindi digit recognition systems that use the same recognition classifier tool especially if we consider the way of writing and number of training and testing samples. In addition, the recognition rate could be increased if we have a high resolution scanner and increase the number of training and testing samples. Key-Words: Offline Hindi Digit Recognition, Hindi Digits, Feature Extraction, MLP, Neural Networks, Backpropagation.