该文提出了一种基于布尔神经网络的数字字符识别的改进算法。该算法以字符的物理结构构成二值特征向量，通过对标准模式样本的训练以及相应吸引半径的初值标定，建立字符识别模型；根据一近邻准则，最大距离准则完成对待识别字符的归属的判决。实验结果表明，该方法具有较强的鲁棒性。 This paper presents an improved algorithm of digital character recognition based on Boolean neural network. The algorithm constructs a binary eigenvector based on the physical structure of the character. By training the standard pattern and calibrating the initial value of the corresponding attracting radius, a character recognition model is established. According to a nearest neighbor criterion and a maximum distance criterion, the judgment of ownership of the recognized characters is completed . Experimental results show that this method is robust.