A Hopfield Based Dynamic Neural Network for Telugu Character Recognition

摘要

Telugu is one of the most predominantly spoken languages in India by millions of people. Not much work has been reported on the developments for handwritten character recognition in most Indian languages. In this paper, it is proposed that a novelistic Hopfield based dynamic neural network should be used to recognize offline hand written Telugu characters. It aims to design a recognizer which manipulates inherent characteristics of the Telugu script. The recognition system comprises two stages-feature extraction and recognition. The first step, feature extraction is accomplished by wavelet multi-resolution analysis and the second step, recognition by associative memory model. The entire proposed system uses Hopfield based dynamic neural networks (DNN) for this purpose. This process overcomes the innate problems of memory limitation and apocryphal states in the Hopfield network. The DNN which has earlier been proved to be efficient for associative memory recall [11] is now being used in the system. The experimental results have been extremely reassuring.