擴充相關LMS之迴音消除

Abstract

對於迴音消除而言，雙邊談話是一個很嚴重的問題，它使得一些採用誤差迴授的適應性濾波器如LMS和RLS無法追蹤房間的脈衝響應。在本篇論文中，我們將使用擴充相關LMS (ECLMS) 演算法來解決雙邊談話的問題，此種演算法是藉輸入信號之相關函數作為系統輸入，而非一般直接以輸入信號的方法。ECLMS演算法有較高的ERLE值，但是所需之計算量也相對的較高，因此我們提出一個簡化的ECLMS演算法來減少所需之計算量。為了讓ECLMS演算法能適用於實際的語音環境，我們提出了正交ECLMS(OECLMS)演算法。其後之理論分析與電腦模擬在在顯示ECLMS與OECLMS有很好的收斂表現甚至在有雙邊談話的情況下亦然。

In the acoustic echo cancellation, double-talk is a serious problem. It makes the echo canceller fail to trace the room impulse response especially for some error feedback adaptive filter like LMS and RLS. In this thesis, we use the Expanded Correlation LMS algorithm to solve the double-talk problem. This algorithm is based on the processing of the correlation functions of the input signal instead of processing of input signal itself. The ECLMS algorithm has high value of the echo return loss enhancement but needs high computational complexity. Therefore, we propose a simplified ECLMS algorithm to reduce the complexity of the ECLMS. To adapt the algorithm to the practical speech implementation, the orthogonal ECLMS was proposed. The theoretical analysis and the computer simulation show that the ECLMS and the OECLMS have good convergence performance to the echo cancellation even the double-talk condition existed.