Handwritten Digit Recognition Based On Prioritized Neural Networks

Handwritten numeral recognition is an important research direction of pattern recognition and image processing.Although simple strokes and only ten kind of handwritten numerals,due to different writing habits and various of glyph,the rate of recognition has not been very ideal.Handwritten numeral recognition system design includes identifier(classifier)design and use of identifier(classifier)method.Shou-jue Wang academician proposed priority sequence of neural network is controllable reject function and higher recognition rate of recognizer.This paper reference the priority order neural network that Shou-jue Wang academician putted forward,designed a kind of based on priority sequence identifier of the neural network handwritten Numbers.On the other hand,optimized the priority sorting algorithm putted forward by Shou-jue Wang,and improved the priority sequence of neural network.With expansion of MNIST database are tested on a variety of recognizer.Experimental results show that the recognition rate of improvement based on priority sorting of the neural network was obviously higher than that of traditional BP network,and the recognition rate is higher than the priority sorting of the neural network that putted forward by Shou-jue Wang academician.