基于人工神经网络的汉字识别方法研究

Artificial neural network based method research of Chinese characters recognition

电子计算机广泛应用于信息处理中,有极强的算术和逻辑运算能力,有极高的运算速度、精确度和可靠度。但是,它的形象思维能力与人相距甚远。如果计算机具备了模式识别能力,人们就可以使用机器来执行感知任务。文章运用了人工神经网络,模式识别的方法及原理,以Matlab软件作为平台来探讨应用神经网络对汉字进行识别。并通过对汉字样本图象采集输入,汉字图象二值化,行字切分,十进制存储等预处理,分别在有、无干扰的情况下对汉字进行识别,从而评价其性能的优劣。

With the fastest operation speed, most remarkable accuracy and reliability, the electronic computer which has the most powerful arithmetic and logical operation abilities is widely applied to manipulate information. However, the ability of its thinking in images is greatly inferior to human beings. If the computer has the ability of pattern recognition, people can use the machine to carry out the task of perception. Applying artificial neural network, the methods and principles of pattern recognition, and using Matlab Software as a platform, this dissertation discusses that application of neural network to identify characters. By collecting and inputting the images of character samples, binary, cutting and dividing characters, handling the decimal scale storage and so forth, we identify characters with or without interference and then I assess the advantages and disadvantages of its performance.