基于方向小波和BP网络的数字识别

A digit recognition method based on directional wavelet and BP neural network

针对目前数字字符识别的研究现状，提出了方向小波变换和BP神经网络相结合的数字字符识别方法．小波变换具有良好的方向特性，根据这个特点能很好的提取数字字符的特征，以此特征作为输入向量，用BP网络进行数字识别．实验结果表明，该方法大大减少了运算数据量，缩短了识别时间，具有较高的识别率，对一些如笔画断裂等典型噪声有很好的识别效果，是一种有效的识别方法．

In this paper, a digits recognition method based on directional wavelet transformation and BP neural network is proposed according to the present research state of figure recognition. The characteristics of figure images can be picked up primely for the favorable directional characteristics of wavelet transformation, the figure images can be recognized by BP neural network, which uses the characteristics as the input vectors. The experimental results show that this method has high recognizing rate, the time has been greatly shortened, and the quantities of data have been cut down. It is effective to recognize those digits whose strokes have ruptured.