时间反转语音掩蔽的语音信号可懂度的客观评价方法

对于开放型办公室语音掩蔽系统性能的评价,语言可懂度是很重要的一个方面,目前通常采取的客观评价方法是STI。将语音信号按一定时间帧长反转后得到的信号我们称为时间反转语音,时间反转语音已被作为有效掩蔽信号之一。虽然对于由平稳噪声掩蔽的语音信号,STI与主观理解的语言可懂度相关性很好。但研究发现STI不适用于估计由时间反转语音掩蔽的语音信号的语言可懂度。文章分析了STI、PESQ及mNCM客观评价方法并进行了实验,实验结果表明,PESQ及mNCM对于由反转语音掩蔽的语音信号仍能较好估计语言可懂度。文章根据客观评价结果,进一步比较了反转语音掩蔽算法的不同参数（反转帧长与信噪比）对于语言可懂度的影响。发现反转帧长的增加和信噪比的降低会导致较低的语言可懂度。

Objective Measurements of Speech Intelligibility for Speech Masked by Time Reversed Speech

Speech intelligibility is an important aspect for evaluating speech masking system in open-plan offices.STI is a common objective measurement so far.Time reversed speech,which is speech reversed according to certain frame length in time domain,became one of effective maskers.Although STI has good correlation with subjective speech intelligibility when masker is steady noise,it is shown in this paper that STI cannot used to predict speech intelligibility for speech masked by time reversed speech.We analyzed STI,PESQ and mNCM.Results showed that PESQ and mNCM can predict speech intelligibility well.We also compared the effects to speech intelligibility of different parameters(reversed frame length and SNR) for speech with time reversed masker,and found increase of reversed frame length and decrease of SNR would lead to poorer speech intelligibility.