用於聲音回音消除之可變步距改良式Gauss-Seidel虛擬均 衡仿射投影演算法

Variable Step-Size Improved Gauss-Seidel Pseudo Proportionate Affine Projection Algorithm for Acoustic Echo Cancellation

屈羽濬(Yu-Jun Chu)

指導教授 ： 簡福榮

國立臺北科技大學/電資學院/電機工程系研究所/碩士(2011年)

摘要

聲音回音的產生，是因為遠端的聲音從近端的揚聲器出來後，在該空間內反
射，之後便進入近端說話者所使用的麥克風所造成。在這篇論文中，提出了兩種回音消除演算法用來預測遠端的回音並且針對其作消除動作，目的是為了要提升語音通話品質｡第一個方法是可變步距的改良後均衡式仿射投影演算法(VSS- IPAP)，第二個方法是可變步距改良式Gauss-Seidel虛擬均衡仿射投影演算法(VSS-IGSPPAP)。而本篇論文在實驗部分是使用四個虛擬的不同稀疏程度的空間環境。至於判定個演算法效能所使用的準則，通常是位錯(Misalignment)(又稱為系統距離(System Distance, SD)以及回音往返耗損增強(ERLE)。在實驗結果部分，本論文會證明所提出的方法優於其他傳統的方法，像是最常見用於回音消除的正規化最小均方演算法(NLMS)。不論如何，與傳統的AP演算法一樣，VSS- IPAP演算法的運算量過於龐大會使得在實用上所需的成本非常昂貴，然而VSS- IGSPPAP演算法就正好相反，該方法不但可以在不同稀疏程度的環境維持
快速的收斂速度，對於運算量的節省，也非常的好(於投影階數為4的時候相較於VSS-IPAP可節省46.68%的乘法運算量)。

關鍵字

可變步距 ； 聲音回音消除 ； 均衡式 ； 改良後均衡式 ； Gauss-Seidel虛擬

並列摘要

Acoustic echo usually arises when sound from a loudspeaker is picked up by the microphone in the same room. In this thesis, two acoustic echo cancellation algorithms are proposed to anticipate the sound signal and cancel it out in order to improve the voice quality on a telephone call. The first one is the variable step size improved proportionate affine projection (VSS-IPAP) algorithm, and the second one is the variable step size improved Gauss-Seidel pseudo proportionate affine projection (VSS-IGSPPAP) algorithm. Four pseudo room space environments are simulated by different sparseness room impulse responses (RIR) in the experiment. The relative system distance (SD) and echo return loss enhancement (ERLE) are often used as the measure criteria to compare different acoustic echo cancellation algorithms. The experimental results show that both proposed algorithms are able to achieve superior performance than the well-known normalized least-mean-square (NLMS) algorithm and other traditional acoustic echo cancellation algorithms. However, like the affine projection (AP) algorithm, the VSS-IPAP algorithm always suffers a large amount of computational load. On the contrary, the proposed VSS-IGSPPAP algorithm not only provides fast convergence speed in different sparseness room impulse response (RIR) situations but also consumes low computational cost.

並列關鍵字

Variable Step-Size ； Acoustic Echo Cancellation ； Proportionate ； Improved Proportionate ； Gauss-Seidel Pseudo

參考文獻

[1] E. Hansler and G. Schmidt, Topics in Acoustic Echo and Noise Control, 1st Ed ; Springer, 2006.

[3] P. S. R. Diniz , Adaptive Filtering, 2nd Ed; Springer, 2008.

[4] P. Vary and R. Martin, Digital Speech Transmission: Enhancement, Coding and Error Concealment; John Wiley & Sons, 2006.

[5] E.A.P. Habets, Room Impulse Response Generator; Department of Electrical and Electronic Engineering Imperial College London, Sept. 2010.

[7] C. Paleologu, J. Benesty, and S.Ciochina, “A Variable Step-Size Affine Projection Algorithm Designed for Acoustic Echo Cancellation,” IEEE Transactions on Audio, Speech, and Language Processing, vol. 16, no. 8, pp. 1466–1478, Nov. 2008.