Handwritten numeral recognition is a branch of character recognition and attracted great attention because of its lesser classification and beneficial demand in practice. Furthermore, with the rapid development of computing technology and image processing by recent years, numeral recognition has found successful applications on business, automatic input and other fields.
People have developed some methods on the research of handwritten numeral recognition nowadays. They are based on neural network, stroke feature extraction, genetic algorithm, wavelet transform, FFT transform, SVM algorithm, template matching algorithm and etc. However, numeral recognition of every kind of font, especially offline numeral recognition has still on its developing stage and the recognition rate is still not very perfect due to the variety of written styles. For these reasons, to find simple and effective methods is still a challenging task. This paper designs a new method of handwritten numeral recognition by digital image preprocessing, feature extraction based on neural network method of judgment and the combined use of Matlab toolbox for the artificial neural network function. Experiments show that the method can gain better recognition rate.