基于功率谱的非均匀子带自适应滤波

Power spectrum based non-uniform subband adaptive filtering

根据输入信号的功率谱特征,非均匀滤波器组能够动态调整子带滤波器的带宽和位置,可以获得比均匀滤波器组更好的收敛性能.根据子带最小均方(least mean square,LMS)自适应算法的收敛特性,以提高其收敛速度为目标,提出了一种基于功率谱信息的非均匀子带自适应滤波方法,并详细讨论了非均匀滤波器组实现结构和信号重建性能.计算机仿真实验和分析表明,与现有的均匀和非均匀滤波器组设计方法相比,在子带数目相同的条件下,本文提出的方法能有效地降低子带信号的最大特征值扩散度,而信号重建性能基本相同.

By utilizing the power spectrum of wideband signals,non-uniform filter banks can dynamically make subband decomposition to provide better convergence performance than uniform ones.A new method for non-uniform subband adaptive filtering was proposed,to improve the convergence speed of the subband least mean square（LMS） adaptation.The implementation structure for this kind of non-uniform filter banks was presented and the signal reconstruction performance was analyzed.Computer simulation results and theoretical analysis show that,compared with existing methods with equal number of subbands,the proposed one can effectively decrease the maximum eigenvalue spread of subband signals while maintaining nearly the same reconstruction performance.