模式识别理论中的逻辑特征方法

本文提出了一个新的模式识别特征方法——逻辑特征方法。这个方法不同于经典的模式识别方法:统计特征方法和句法结构方法。在逻辑特征法中,决定模式特征匹配状态的只有“是”或“非”,而不象统计特征和句法特征那样有匹配程度之分如“远”或“近”。逻辑特征有两个重要特性。1.多个逻辑特征可以链接在一起做为逻辑特征向量,这个逻辑特征向量可以做为索引匹配的索引号而对模式进行直接和索引分类。2.逻辑特征还可与其他特征共同使用,其中逻辑特征将做为标志。亦即只有逻辑特征匹配了,才能用其他特征进行匹配程度的计算。

LOGICAL FEATURE METHOD IN PATTERN RECOGNITION THEORY

Logical feature method--a new method in pattern recognition theory--is proposed in this pa-per. The method is diffrtent from the classical pattern recognition methods--statistical featuremethod and syntax feature method. When determining the match situation in the logical feature method, there is only YES or NOTinstead of near or far. The logical feature has the following two properties: 1. Many logical features can be linked together as a logical feature vector. The logical fea-ture vector can be used as index number to perform index matching. 2. The logical feature is also used with other features. When it is used with other features,the logical feature is used as flag, In the classification procedure, the other features can be used if andonly if the logical feature is matched.