本文采用前向、多层神经网络,BP学习算法对40个人的手写体数字进行了识别。识别过程分为四步:首先,用HP扫描仪把写在纸上的数字变成二值图像,接着对它进行分割,规整等预处理,变换成32×32点阵。然后提取特征,把点阵图像变成特征描述。最后,进行训练和识别。在拒识率为25％条件下,得到误识率为0.4％的识别结果,文中还分析和讨论了在实验中遇到的一些问题。 In this paper, 40 human handwritten numbers are identified by using forward, multi-layer neural network and BP learning algorithm. The recognition process is divided into four steps: First, a digital image written on paper is converted into a binary image by an HP scanner, and then the image is divided and preprocessed to be a 32 × 32 dot matrix. Then extract the features, the bitmap into a feature description. Finally, training and identification. When the reject rate is 25%, the recognition result with the error rate of 0.4% is obtained. Some problems encountered in the experiment are also analyzed and discussed.