本文以音节作为声调识别基元,以简化FUZZY ARTMAP作为声调识别系统模型,研究了具有连续语音音节特征的三字词的声调识别问题,在此基础上,对三字词进行了声调识别规则的自动抽取,经网络剪枝、权值量化和规则合并后,得到了若干条与声调模式的统计分布一致的规则,这些规则具有一定的普遍意义,可用于声调识别和声调模式分析,获取规则的方法可直接应用于连续语音。 In this paper, we use the syllable as the tone recognition primitive to simplify the FUZZY ARTMAP as the tone recognition system model, and study the tone recognition of the three words with continuous speech syllables. On this basis, Automatic extraction, after the network pruning, weight quantification and rule combination, obtained a number of rules consistent with the statistical distribution of tone patterns, these rules have a certain universal meaning, can be used for tone recognition and tone pattern analysis, access to the rules The method can be applied directly to continuous speech.