A new approach for non-uniform subband adaptive filtering

In this paper, we propose a non-uniform subband adaptive filtering (SAF) method using a new objective. This objective uses the decrease in power in each subband error signal. Carrier frequencies, decimation rates (parameters in the analysis bank), lengths of each adaptive filter and FFT bin number are treated as time-varying parameters, and adjusted to ensure quick convergence of subbands with small power decrease, by sacrificing convergence of subbands in whose power decrease is large. The proposed method uses a delayless structure to avoid signal path delay. Simulation with colored noise shows that our method can achieve higher convergence rates and smaller MSE reductions than uniform SAF.