普通话发音质量评价算法改进研究

Improvement in automatic putonghua pronunciation quality assessment algorithm

隐马尔科夫模型的对数后验概率算法是计算机辅助语言学习系统中用来衡量考生发音质量的重要指标。但在普通话智能测试系统中,传统的后验概率算法与专家评分之间存在比较明显的差距。文章从普通话语音评价的主观标准出发,将普通话发音的语言学知识引入后验概率算法,重构算法的语音识别网络,同时从音素评分模型角度对现有的发音质量评价算法进行改进。

HMM based log posterior probability is an important feature for computers to judge the test- er＇s pronunciation quality in computer assisted language learning systems. However, the discrepancy between traditional posterior and evaluator＇ s criteria is obvious in the Putonghua proficiency test. In the paper, we introduced the linguistic knowledge of Putonghua into the log posterior algorithm aiming at the subjective standard Putonghua speech evaluation, reconstructed the speech recognition network, and improved the existing Putonghua quality assessment algorithm from the score model.