维吾尔语电话语音语料库的研发

Development of Uighur Speech Phone Corpus

本文根据现有的维吾尔文语音识别语音库的不足,以自然口语为对象研究维吾尔语的语音特征,提出了适合该语言的电话语音语料库设计方案,其中包括了维吾尔语电话语音库的文本设计、发音人的选择、语音录制、语音库的标注和后期处理方法等.本文从构建的350个说话人的维吾尔语电话语音语料库中挑选50个目标人提供给基于GMM-UBM/SVM的维吾尔语电话信道说话人识别的研究.

This article focuses on the research of the phonetic features in Uighur language in consideration of the shortage of currently Uighur corpus. An appropriate design of Uighur phone speech corpuses is proposed to settle the problems in Uighur text designation, speakers selection, voice recording, annotation, and post-processing methods. An Uighur phone speech corpus of 350 speakers is finally constructed, and a sub-corpus of 50 speakers is used in the research of GMM-UBM/SVM based speaker identification under the Uighur phone channel.