High Accury Handwriting Digit Recongnition Sysytem

In this paper, a system based on two grade BP neural network is set up. During the implementing of this handwriting digit recognition system, the main work is done as following:1. 109 dimensions features of Seven types for handwritten digits based on different considerations are extracted, which are applied the two grade BP neural network system.2. Make some improvements on BP neural network to quicken the network constringency speed and to avoid fake saturation phenomenon. For example, change the learning-factor.3. Analyzed the shortcomings of one grade BP neural network, and by a great deal of statistic work, suggested to implement the handwriting digit recognition system with two grade BP neural network. And based on the statistic results designed the second grade of BP neural network. Illustrated the advantage of the two grade neural network system.4. The result of the experiment show that the two grade neural network system can get higher correct rate at a certain extent than the one grade BP neural network.