中文摘要：

湘西方块苗文是湖南省湘西少数民族地区使用的一种文字,在采用传统的BP神经网络对字符图像进行识别时,易出现收敛速度慢、陷入局部极小值的问题。对此,为采用BP神经网络对湘西方块苗文图像进行更好地识别,提出了一种基于附加动量法和自适应学习速率相结合的改进型BP神经网络,并将其应用于湘西方块苗文的识别。仿真结果表明,该算法不但收敛速度快,而且可以避免陷入局部极小值。

英文摘要：

Square Hmong Language Characters of Xiangxi is used in Hmong area of Xiangxi. The images of characters recognition based on traditional BP neural network can easily fall into local minimum and slow convergence. As to overcome this drawbacks, improved BP neural network based on integrating the additional momentum method and adaptive learning rate methods was presented by this article. This algorithm could be used to recognize the images of Xiangxi Square Hmong Language Characters. The simulation shows that not only could the algorithm improve the convergence speed, but also get rid of local minimum.