详细分析了语音评测中的基本策略——对数后验概率评价策略,提出利用语音自身的特征相似性组织判决音素集的方法,既提高了评测精度又减少了计算量.在分析普通话声母显著性特征的基础上,提出了基于预分类的评测方法.同时,根据音征互载的特点,提出了声韵结合和韵声韵结合的音素模型.声母评测系统取得83.56%的评测准确率,基本达到人工评测水平. This paper analyzes in detail the basic strategy of voice evaluation - the logarithm posterior probability evaluation strategy, puts forward the method of using phonemes’ own feature similarity to judge the phoneme set, not only improves the evaluation accuracy but also reduces the computational cost. Based on the characteristics of the sound, a pre-classification based evaluation method is proposed.At the same time, according to the characteristics of the interaction of sound and sound, a phoneme model combining rhyme and rhyme rhyme is proposed.The evaluation accuracy of 83.56% Artificial evaluation level.