并联式多级回声抵消器设计研究  

Study of paralleled multi-channel acoustic echo canceller design

作　　者：孟翼[1] 邓北星[2] 黄永峰[2] 

机构地区：[1]中国科学院研究生院,北京100049 [2]清华大学电子工程系,北京100084

出　　处：《电子测量技术》2007年第10期119-121,共3页Electronic Measurement Technology

摘　　要：在回声抵消研究中,LMS算法由于具有算法简单、易于实现的特点而得到广泛应用。然而,采用LMS算法的自适应回声抵消器的性能仍然有很大的改进余地,可以通过对算法本身的改进和对利用该算法的自适应滤波器的结构的改进达到更好的性能,主要研究了利用LMS算法的AEC自适应滤波器结构上的改进,提出了一种新的滤波器设计方法,该方法通过将采样信号在输入端拆分,分别输入到多个并联的子滤波器中分别处理,得到对回声信号的估计后将它从掺杂有回声信号的接收信号中消除,并在输出端重新组合,成为对原始信号的恢复。这种并联多通道结构可以有效提高回声抵消器的收敛速度、降低运算量。由于这种方法利用了多通道并联自适应滤波器的结构,因此要求非常严格的数据同步和时间控制。仿真结果表明,这种方法对回声抵消具有良好的效果。In the study of acoustic echo cancellation, LMS algorithm is widely used because of its good robustness and simplicity in algorithm structure and implementation. However, there are still plenty of spaces for the LMS AEC filter performance to improve its adaptive acoustic echo cancelling performance either from the algorithm itself or from the structure of adaptive filter. This article is mainly focus on the better design structure of AEC with LMS algorithm. The paper proposes a new solution for AEC with a new structure get further improvement in AEC performance. In the proposed method, sampled voice signal is divided into several pieces as an input for a paralleled multi-channel adaptive filter and stored in the register in each sub-channel. The signal in each channel is handled by each sub-filter separately. An estimated echo signal is supposed to be generated. The combination of the estimated echo signal will be subtracted at the end of the output point of the multi-channel adaptive filter, an estimated voice signal will be recovered. The new designed multi-channel adaptive AEC has a better performance with higher speed and lower calculation load. To achieve expected performance, timing and synchronization are strictly required. Simulation proves that this parallel multichannel adaptive AEC can meet the expected effect.

关 键 词：自适应滤波 回声抵消器 多级抵消 寄存器 同步 

分 类 号：TN912.3[电子电信—通信与信息系统]