基于语音识别和语速修改的语音复读系统  

Speech Repeating System Based on Speech Recognition and Speaking Rate Modification

作　　者：梁青青[1] 杨鸿武[1] 郭威彤[1] 裴东[1] 

机构地区：[1]西北师范大学物理与电子工程学院,兰州730070

出　　处：《计算机工程》2011年第5期288-290,共3页Computer Engineering

基　　金：国家自然科学基金资助面上项目(60875015);教育部科学研究重点基金资助项目(208146)

摘　　要：针对英语学习中的听力练习问题,利用语速修改算法和大词表连续语音识别算法,实现一个面向英语学习的语速可变、字幕同步的数字复读系统,根据字幕选择相应的语音进行复读,并实时调整语速。MOS评测结果表明,系统调节语速后的语音平均MOS得分为4.1,接近原始语音质量。语音识别结果显示,系统对英语听力材料中纯净语音的识别率达到70.8%,能够满足英语听力学习的需要。Aiming at the problem of listening exercise in English learning,this paper proposes a speech repeating system which adjusts the speaking rate with TD-PSOLA algorithm and can display the subtitle of speech by large vocabulary connected speech recognition method.With the system,users can select the speech to repeat by selecting subtitle and modify speaking rate of selected speech real-time.The modified speech by system achieve 4.1 of the average Mean Opinion Score（MOS）,which is close to the quality of the original voice.Result of speech recognition evaluation shows that the word level accuracy of speech recognition on pure English learning material is 70.8%.

关 键 词：语速调节 语音识别 字幕同步 语音复读 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]