一种计算复杂度低的双端通话检测算法

Low computational complexity double-talk detection algorithm

在声学回音消除中,近端语音的出现会导致模拟回音路径的自适应滤波器发散,一个成熟的声学回音消除器应该包含有双端通话检测算法。针对这个问题,提出了一种计算复杂度低的、基于麦克风信号与误差信号的互相关双端通话检测算法。同时,该算法与滤波器系数缓存机制相结合以进一步提高系统的鲁棒性。实验结果表明,该算法具有良好的检测性能,可以对双端通话的出现和消失做出快速响应,同时能显著提高系统在双端通话环境下的回音消除效果。

In acoustic echo cancellation,the presence of near-end speech causes divergence of the adaptive filter which is used to model the echo path.A robust acoustic echo canceller must be equipped with a double-talk detection algorithm.This paper proposes a novel normalized double-talk detection algorithm based on the cross-correlation between the microphone signal and the cancellation error.The algorithm is designed in such a way that it meets the needs of a low computational complexity double-talk detector.In addition,the algorithm is coupled with coefficient-memory in order to improve its robustness.Simulation results demonstrate that the system can quickly react to onset and cessation of double talk detection and significantly improves echo cancellation performance in double-talk situation.