聲訊迴聲消除之研究

Abstract

本論文主要是對迴聲消除器做整體的介紹，並探討數種適應性演算法在 回聲消除器上的表現。一般的聲訊迴聲消除器若要以有限響應濾波器實現 ，則其係數經常需要數百個或是數千個，硬體的運算量十分龐大，而且收 斂的速度也很緩慢。在本文中，我們探討各種演算法的收斂速度，以及迴 聲消除的效果，並針對收斂緩慢的特性提出了外插式演算法來加快收斂速 度。最後利用計算機輔助設計工具(DSP Station)從事硬體的實作。 In this thesis, the theories of echo cancellor were studied,and the performance in terms of speed, stability and complexityof several adaptive algorithms applied to echo cancellor wereevaluated. Generally speaking, for implementing an echo cancellor,we need an FIR with hundreds and even thousands of taps. Thecomputation complexity and convergence time are huge accordingly.For solving the slow convergence problem, we propose a linearextrapolated algorithm which can improve the speed of convergence.An implementation is then given with the help of CAD tools(DSP Station).