汉字识别系统的误识模型  

Mis recognition model of Chinese character recognition system

作　　者：马少平[1] 夏莹[1] 朱小燕[1] 姜哲[1] 

机构地区：[1]清华大学计算机科学与技术系

出　　处：《清华大学学报（自然科学版）》1998年第S1期111-114,共4页Journal of Tsinghua University(Science and Technology)

基　　金：国家"八六三"高技术项目;国家自然科学基金

摘　　要：为评价一个汉字识别系统的可信度，提出了汉字识别系统误识模型的概念，给出了误识模型的定义。在实验分析的基础上，确立了影响系统识别性能的两个主要参数：首选距离及其与二选的距离差。给出了运用统计分析方法建立识别系统误识模型的方法。对一个现有的手写体汉字识别系统，经过对其识别性能的统计分析，建立了一个误识模型，初步测试表明了该方法的可行性，达到了实用的要求。误识模型可用于识别率的估计、拒识处理、后处理、多识别器集成和综合评判等多个方面。In order to resolve how to evaluate the reliability of a recognition system, the researchers concept of mis recognition model of Chinese character recognition system, and the definition of mis recognition model are presented. Then on the basis of experimental analysis, two main parameters are ascertained, that have great effects on the recognition performance of the system: the first condidate's distance and the distance difference between it and the second candidate. A method to establish mis recognition model through statistical analyses and calculations is given. A mis recognition model, through the statistical analysis on the recognition performance of a nowavailable handwritten Chinese character recognition system is established. Elemental testing shows the feasibility of this method that can fill the requires of application. Mis recognition model can be used in estimate of recognition rate, refusal processing, post processing, multi recognizer integration and synthesized judgement.

关 键 词：汉字识别 误识模型 后处理 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]