论文部分内容阅读
利用神经网络技术对无限制手写体数字进行识别。在选取合适的特征基础上,对图像进行分域处理,每一小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。
Using Neural Network Technology to Identify Unlimited Handwritten Digits. On the basis of selecting suitable features, the image is divided into regions, each small region corresponds to a neuron in the input layer. The test results show that the identification scheme has strong anti-distortion, anti-rotation ability.