摘要
构建手写数字识别系统,并对系统中BP网络反传学习速率进行改进,提高了识别率并缩短了学习时间;在特征提取上针对数字综合采用二值化并统一规格化的方法,优化了系统的识别精度并提高了识别速度。采用BP算法增强了数字识别的容错性、鲁棒性。
The construction of Handwritten Numeral recognition system,and the recognition rate is improved。 Learning time is also decreased by improving the Back Propagation（BP） network.On the feature extraction,Numerals use two value and unite standard method,These feature extraction methods optimize the system and improve the accuracy of the identification of the speed.By using BP algorithm,fault-tolerant and robust of numeral recognition are enhanced.
关键词
数字识别技术
BP算法
特征提取
神经网络
Numeral recognition
BP algorithm
feature extraction
neural network
