张正文 饶鼎 巩朋成 涂斯纯

摘  要： 声学回声消除（AEC）系统依赖于自适应滤波器的回声路径估计，当麦克风接收的信号存在回声和近端信号时，可能出现模拟回声路径的自适应滤波器发散，导致回声消除性能下降，严重时影响双端通话质量。一个成熟的声学回声消除器应该包含有双端通话检测算法，针对该问题，提出一种准确度高、性能稳定的基于信号包络检测（Env）和归一化互相关（NCC）估计相结合的双端通话检测（DTD）算法。该算法先使用远端信号与麦克风信号的能量包络来判断双端通话是否发生，当包络检测不能准确地判断双端通话时，再使用远端信号与麦克风信号的互相关估计对双端通话进行最终判断，保证双端通话检测的准确性。仿真实验表明，提出的算法可以准确判断双端通话的开始与结束，并且受回声路径变化的影响较小，提升了AEC系统的性能。

关键词： 双端通话检测; 声学回声消除; 自适应滤波; 回声路径估计; 包络检测; 互相关

中图分类号： TN912.3?34                        文献标识码： A                         文章编号： 1004?373X（2020）03?0022?05

Double?talk detection algorithm based on signal envelope and cross correlation

ZHANG Zhengwen， RAO Ding， GONG Pengcheng， TU Sichun

（School of Electrical and Electronic Engineering， Hubei University of Technology， Wuhan 430068， China）

Abstract： The acoustic echo cancellation （AEC） system relies on the echo path estimation of the adaptive filter. When the signal received by microphone has echo and near?end signal， the adaptive filter divergence of the analog echo path may occur， resulting in degraded echo cancellation performance， which will affect the double?talk quality in severe cases. A mature acoustic echo canceller should be based on double?talk detection algorithm. In view of this， an accurate and stable double?talk detection （DTD） algorithm based on the combination of signal envelope detection （Env） and normalized cross correlation （NCC） estimation is proposed. In the algorithm， the energy envelope of the far?end signal and the microphone signal is used to judge whether the double?talk occurs or not. When the envelope detection fails to judge double?talk accurately， cross?correlation estimation of the far?end signal and the microphone signal is adopted to perform the final judgment of the double?talk to ensure its accuracy. The simulation experiments show that the proposed algorithm can accurately judge the beginning and end of double?talk， and is less affected by the change of echo path， which improves the performance of AEC system.

Keywords： double?talk detection; acoustic echo cancellation; adaptive filtering; echo path estimation; envelope detection; cross correlation

0  引  言

在視频会议系统、IP对讲通话系统中，扬声器播放的声音经过多种路径传播后被麦克风拾取，并传回到远端形成声学回声，影响通话质量。因此，可以使用声学回声消除（Acoustic Echo Cancellation，AEC）技术来消除声学回声。

在声学回声消除系统中，自适应滤波器用于消除远端语音的回声，任何近端信号的存在都会影响其收敛[1?2]。在双端通话期间，由于远端语音信号与近端语音信号的不相关特性，近端信号表现为一个很强的干扰信号，使得自适应滤波算法性能下降，甚至出现自适应滤波器发散，导致讲话人听到自己的回声，这是非常糟糕的体验[3?4]。……