论文部分内容阅读
双端语音检测是声回波对消系统的重要组成部分。鉴于目前采用的基于相关的双端语音检测运算量较大,文中采用基于相干的双端语音检测方法,利用多窗谱估计法估计相干函数。该谱估计法能够利用较短的数据段获得可接受的频率分辨率而估计方差较小。软件仿真结果表明,基于相干的双端语音检测与基于相关的双端语音检测相比较,前者能够在不严重增大漏报概率的情况下大大提高运算效率。
Double-ended speech detection is an important part of the acoustic echo cancellation system. In view of the large amount of computation based on the related double-side speech detection, the coherent double-side speech detection method is adopted in this paper, and the coherence function is estimated by using the multi-window estimation method. The spectral estimation method can use shorter data segments to obtain acceptable frequency resolution and estimate the variance less. Software simulation results show that the coherent-based double-ended speech detection can greatly improve the computational efficiency without seriously increasing the underreporting probability compared with the correlation-based double-ended speech detection.