“人工嘴”在语音声学分析中的应用研究

Mouth Simulator in Acoustic Analysis

"人工嘴"是一种可以模拟人嘴发声的放音装置。它在近场范围内能够精确模拟人嘴所产生的声场,专门用于电话、手机及电声元器件等传输和通信类产品中麦克风通路的声学特性的检测。本文主要介绍通过提取"人工嘴"播放的语音(测试语音)与源语音(音源)的语音质量感知评估(perceptual evaluation of speech quality, PESQ)得分、基频和能量、共振峰、感知线性预测系数和信道因子等五个声学参数,分析比较二者之间存在的差异,来评估"人工嘴"对语音声学特征的影响。通过实验可知,"人工嘴"作为一种新型的放音装置,在使用过程中对语音质量和语音声学参数的影响是客观存在的,但程度不同;差异最大值出现在PESQ的评测结果中。这为"人工嘴"在今后相关项目中的应用奠定了数据基础,具有一定的参考价值。

Mouth simulator,an audio device used to imitate human voice,can exactly mimic human mouth speaking to generate sound field in near range,having its special purpose for either sound transmission through telephone,mobile phone,acoustic components or detection of acoustic features about microphone channel in communication products.With the extracted voices of the sampling(broadcast by mouth simulator for test)and the source,the comparison was here made on five acoustic parameters of PESQ(perceptual evaluation of speech quality),fundamental frequency,formant,PLP(perceptual linear predictive)and i-vector so that the difference can be recognized between the sampling and source voice,making the impact assessed of mouth simulator on acoustic features.The experiment indicated that mouth simulator,as a new type of audio device,inevitably has its effect on speech quality and acoustic parameters in practical usage despite the different degree.PESQ evaluation showed the maximum difference.