语音识别中声效模式的分析及检测.pdfVIP

优先出版 计 算 机 应 用 研 究 第32 卷 语音识别中声效模式的分析及检测 晁 浩，宋 成，彭维平 (河南理工大学 计算机科学与技术学院，河南 焦作 454000) 摘 要：针对声效相关的语音识别鲁棒性问题，在分析了声效变化情况下声强、时长、帧能量分布以及频谱倾斜能方面 特性的基础上，建立了基于GMM 的声效检测器。同时，还研究了声效变化对语音识别精度的影响，并提出了基于多模 型框架的语音识别算法。汉语孤立词语音识别实验显示，除正常模式的语音识别精度略有下降外，其它四种声效模式的 识别精度均有大幅度的提高。实验结果表明语音信号的声强、时长、帧能量分布以及频谱倾斜等信息能够用于识别声效 模式，而多模型框架是解决声效相关的语音识别鲁棒性问题的有效方法。 关键词：语音识别；声效；多模型框架；高斯混合模型；孤立词 中图分类号：TP391.42 Analysis and detection of vocal effort for speech recognition CHAO Hao, SONG Ceng, PENG Wei-ping (School of Computer Science Technology, Henan Polytechnic University, Jiaozuo Henan 454000) Abstract: The impact of changes in a speaker’svocal effort on the performance of automatic speech recognition is the research target of the paper. Firstly, Analysis is performed for sound intensity level, duration, frame energy distribution and spectral tilt, and vocal effort detector based on GMM is proposed by using these phonetic properties and acoustic properties. Then, Impact of vocal effort variability on automatic speech recognition is studied, and multiple model framework approach is presented to improve robustness of speech recognition system. Experiments conducted on isolated-word recognizer show that accompanied by a slight decline for normal mode, significantly improvement of recognition accuracy for the remaining four vocal effort modes can be achieved. Thus, potential of the method is demonstrated. Key Words: speech recognition; vocal effort; multiple model framework; Gaussian mixture model; isolated-word 特性的变化规律并消除这种变化对语音识别产生的消极影响， 0 引言