模式识别统一熵理论

Information Entropy Theory in Pattern Recognition

本文提出的模式识别统一熵理论包括;对模式识别中存在着两种信息源的分析、利用信息熵概念定义模式识别信息熵系统、论述学习和识别过程中信息熵的变化和传递以及提出模式识别信息熵三定理.统一熵理论揭示了模式识别信息过程中特征选择和模式识别性能极限间的内在连系,对模式识别理论研究和系统设计都具有一定指导意义,这一理论在汉字识别的研究和实践中得到了验证.

A Pattern Recognition Information Entropy Theory (PRIET) is presented in this paper. Two kinds of information sources in pattern recognition are analysed,the information entropy system in pattern recognition is defined,using the concept of information entropy,the convertion and transition of information entropies in learning and recognizing process are discussed and three information entropy theorems are presented. The PRIET reveals the inherent relationship between the feature selection and PR performance limitation,and is helpful for researching and designing the PR system. The PRIET is proved by the practice of Chinese character recognition system.