基于多Agent系统的脱机手写体汉字识别  

Offline recognition of hand-written Chinese characters based on a multi-Agent system

作　　者：马少平[1,2,3] 金奕江[1,2,3] 

机构地区：[1]清华大学计算机科学与技术系,北京100084 [2]清华大学智能技术与系统国家重点实验室,北京100084 [3]清华大学清华信息科学与技术国家实验室(筹),北京100084

出　　处：《智能系统学报》2009年第5期398-405,共8页CAAI Transactions on Intelligent Systems

基　　金：国家自然科学基金创新研究群体科学基金资助项目(60621062)

摘　　要：由于脱机手写体汉字的多样性和随意性,识别起来具有很大的难度,依靠单一的特征很难实现高准确率的识别.引入多Agent的概念,将多种知识统一于多Agent系统之中,给出了一个面向脱机手写体汉字识别的多Agent类市场模型,提出了一种模糊综合方法和辩论协商规则,实现了一个基于多Agent系统的脱机手写体汉字识别系统.初步测试结果显示出系统的有效性.Due to the diversity and randomness of Chinese characters, it is difficult for offline hand-written Chinese character recognition to perform well when based solely on analysis of a single feature. In order to solve this problem, a multi-Agent based recognition method was proposed. It merges a variety of knowledge into a market-like model. A comprehensive approach using fuzzy rules to provide consultation and debate rules between Agents was also incorporated. With this proposed method, a multi-Agent offline hand-written Chinese character recognition system was constructed. Preliminary experimental results showed the effectiveness of this system.

关 键 词：汉字识别 多AGENT系统 类市场模型 模糊综合 辩论协商规则 

分 类 号：TP391.43[自动化与计算机技术—计算机应用技术]