采用预测残差的回声消除单双端检测算法

Echo Cancellation Single and Double-Talk Detection Algorithm Using Predictive Residuals

在单端与双端状态下,自适应滤波器的预测残差数值迥异,对预测残差数值进行归一化处理,拉开单端、双端状态的数值差距,扩大阈值范围.提出基于预测残差的回声消除单双端判断方法,并提出在单双端交界处使用自适应滤波器权值复用与更新机制,以及算法的实现步骤.实验结果表明:该算法具有宽广的阈值选择范围,可以有效防止因单双端误判引起的滤波器发散,减小总的回声消除预测误差.

In the condition of single-talk and double-talk,the predictive residual data of self-adapting filter is different.this article set normalization processing in predictive residual data,opened numerical gap in singletalk and double-talk,and expanded the scope of the threshold value,then this article put forward the echo cancellation single and double-talk judgment method based on predictive residuals.And put forward that it can use the adaptive filter working mechanism of weight reuse and update in single-talk and double-talk junction.At last,the implementation steps of the algorithm are also given.The experimental results showed that this algorithm has a wide range of threshold selection,it can effectively prevent the filter divergence caused by inaccurate double-talk detection and reduce the total echo cancellation prediction error.