基于并行神经网络的汉字识别系统

Chinese Character Recognition Based on Parallel Neural Network

介绍了一种基于并行神经网络的汉字识别系统。该系统进行汉字图象的预处理后 ,针对汉字平移、旋转、尺度变化 ,提取三类相对稳定且抗噪、反映汉字结构信息的统计特征作为神经网络的输入。神经网络采用叠层BP网 ,用BP算法进行训练、学习和识别。本系统对标准BP算法做了若干改进 ,从速度和识别率上都得到了明显的提高 ;用PVM网络并行平台虚拟成并行机 。

Introduces an Chinese character recognizing system based on parallel neural network. After pre processing of the image, extracts three sorts of statistical features which are relatively stable ,anti jamming and reflecting structure information .They are input of neural network which is a multilayer BP network topology. Using BP algorithm for learning and recognition, in this system, there are several great improvement. The results are satisfied. PVM platform is used for realizing parallel processing of network.