基于认识的神经网络模型.pdf

基于认知的神经网络模型 摘 要 车牌汉字识别是智能交通运输系统的一个重要组成部分，同时，汉字识别也是汉语、汉 字认知研究的一个重要领域。该文在 1982 年芬兰学者 Kohonen 提出了自组织神经特征映射 神经网络（Self Organizing Feature Mapping，SOM ）的基础上实现了对车牌汉字的聚类分别， 构建一个 SOM 网络的汉字自组织神经网络模型。从汉字字形的聚类和汉字部件的角度，对 基于汉字认知的汉字识别过程进行了初步的探索。模拟研究表明，模型通过训练学习能够识 别出汉字的结构类型和部件，发现汉字识别的规律，在一定程度上模拟了汉字的认知，说明 了基于自组织神经网络用于车牌汉字的认识乃至汉语言学习的可行性。 关键词：自组织神经网络，汉字认知，聚类，汉字识别 I 基于认知的神经网络模型 Abstract License plate character recognition is an important component of Intelligent Transport Systems (ITS), meantime, Chinese character recognition is an important field in Chinese characters cognition. This paper on the basis of the neural network Self Organizing Feature Mapping that a Finland scholar Kohonen posed in1982, build a network of Chinese characters SOM self-organization neural network model. The process of Chinese characters recognition based on Chinese characters cognition is researched from the aspect of Chinese characters cluster and component. Result from this simulation suggests that the model is able to recognize the architecture and components of Chinese characters. It can detect rules of Chinese characters recognition by learning. So it can simulate the process of Chinese characters recognition to some extent, which Based on the self-organizing neural network for the license plates of Chinese characters and Chinese language and understanding of the feasibility study. Key words ：Self Organizing Feature Mapping，Chinese characters cognition, Cluster, Chinese characters recognition