应用BP神经网络进行手写体字母数字识别

Recognize Unconstrained Handwritten Characters in BP Neural Network

该文提出了一种采用BP神经网络方法来对无限制手写体字母、数字进行识别的方案。在选取合适的特征点的基础上,对手写体图像进行分域处理,每一小区域对应输入层的一个神经元。测试结果表明,该识别方案具有很强的抗畸变、抗旋转能力。

This paper introduces neural network to recognize unconstrained handwritten characters.On the basis of appropriate feature extraction, handwritten image segmentations are made.Every part corresponds to a neural cell of input layer.Experimental results show that this method has the ability of rotational variance-resistance.