摘要
目前最小均方(LMS)自适应算法已经成为IP电话回声消除器采用的主要算法。但在一些特别的场合,在IP会议电话中,用于回波抵消的自适应滤波器的阶数非常高,要实时实现自适应回波抵消是非常困难的。介绍了一种归一化快速块最小均方算法(NFBLMS),相比LMS算法,它不仅可以有效减小运算复杂度,同时也能改善收敛速度。最后在MATLAB中生成符合G.168要求的带限复合信号源作为远端参考信号。并运用MATLAB对算法进行仿真,仿真结果表明NFBLMS算法的各项性能指标要优于LMS算法,从而适用于IP会议电话的要求。
Currently,LMS adaptive algorithm has become a primary algorithm adopted by echo canceller of VoIP.But on some special occasions such as conference telephone of IP,the order of adaptive filter is extremely high,in such a case,it is very difficult to implement real-time echo cancellation.The paper introduces a normalized fast least mean square(NFBLMS) algorithm that can reduce complexities of algorithm and optimize the convergence rate in comparison with LMS algorithm.Finally,bandlimited composite source signals(CSS) in line with G.168 will be generated in MATLAB to recognize far-end reference signal.The algorithm is simulated by using MATLAB,the result of simulation indicates that some performances of NFBLMS algorithm are better than that of LMS algorithm,thus,it is applicable to conference telephone of IP.
出处
《计算机仿真》
CSCD
北大核心
2010年第1期322-325,共4页
Computer Simulation
关键词
复合信号源
回声路径
最小均方算法
收敛速度
回声返回损耗增益
Composite source signals
Echo path
Least mean square algorithm
Convergence rate
Echo return loss enhancement