基于HMM的藏语安多方言文本分析研究  

Research on Text Analysis of Tibetan Amdo Dialect Based on HMM

作　　者：苏丽 SU Li(Shandong Vocational University of Foreign Affairs,Rushan County,Weihai City,Shandong Province,264504,China)

机构地区：[1]山东外事职业大学,山东威海264504

出　　处：《新一代信息技术》2020年第23期35-40,共6页New Generation of Information Technology

基　　金：山东外事职业大学校级项目(项目编号:2019YB05)。

摘　　要：隐马尔可夫模型(Hidden Markov Model,HMM)在汉语语音合成中主要用于文本分析,因为文本分析对合成语音的自然度和清晰度有很大影响。语音的训练和合成都要经文本分析阶段,然而目前还没有藏语安多方言的文本分析研究。将HMM方法应用于藏语安多方言的文本分析,获取文本的单因素和上下文标注信息,实现机器的自动标注,具有创新性。实验测评结果表明,基于HMM的藏语安多方言语音合成效果优良。Hidden Markov model(Hidden Markov Model,HMM)is mainly used in text analysis in Chinese speech synthesis,because text analysis has great influence on the naturalness and clarity of synthesized speech.Phonetic training and synthesis must go through the stage of text analysis,but there is currently no research on Tibetan Ando dialect text analysis.This paper applies HMM method to the text analysis of Tibetan Ando dialect,obtains the information of single factor and context annotation of the text,and realizes the automatic annotation of the machine,which is innovative.The experimental results show that the speech synthesis effect of Tibetan Ando dialect based on HMM is good.

关 键 词：安多方言 HMM语音合成 单因素标注 上下文相关标注 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]