摘要
由于BP神经网络具有并行处理信息、自组织、自学习信息等优点,本文采用了BP神经网络对手写数字识别进行运算,提取笔画密度、长宽比和欧拉数等特征作为训练样本。并用Matlab对其算法进行仿真,并且很准确的识别出来,说明其有非常广泛的前景。
BP neural network has predominances of collateral processing information, self-organizing, self-learning. The paper uses BP neural network to recognize handwritten digit samples, and feature of the stroke density, length width ratio and Euler number, and is extracted as training samples. The algorithm is simulated with Matlab. It has a very accurate identification and promising prospect.
出处
《自动化技术与应用》
2014年第5期5-10,共6页
Techniques of Automation and Applications
关键词
BP神经网络
手写数字识别
模式识别
特征提取
BP neural network
handwritten numeral
pattern recognition
feature subset selection
