Investigation Of The Performance Of Frequency Domain Active Noise Control Algorithms

Active noise control (ANC) technology has obtained increasingly wide application in solving the problem of noise pollution. The performance of the commonly used feedforward ANC system is largely determined by the convergence behavior of the adaptive algorithm. Compared with time-domain adaptive algorithms, frequency-domain adaptive algorithms have the potential benefits of fast convergence and low computational burden, which makes them a good choice in the implementation of active noise control systems.Block LMS algorithm (BLMS), Frequency domain LMS algorithm (FBLMS) and Multi-delay block frequency domain (MDF) algorithm are introduced firstly. FBLMS is a fast frequency domain implementation of the BLMS algorithm. FBLMS’ convergence rate can be improved by normalizing the energy of each frequency bin (NFBLMS). The steady-state LMS error is the output of a time invariant linear system which is driven by the wide sense stationary Wiener error. The transfer function mainly depends on the power spectrum estimation of reference signal.Secondly, the convergence behavior of FBLMS algorithm in non-causal conditions is analyzed. It is proved that the steady state solution of the NFBLMS algorithm depends on the variation of the power spectrum estimation. The flatter the reference signal power spectrum is, the better the steady state solution. The corresponding noteworthy deduction is that the steady state solution of the NFBLMS algorithm may become worse with longer filter length. Simulations and experiments method is proposed and analyzed, which can guarantee the optimal steady solution of the frequency domain algorithm. Both simulations and experiments demonstrate the benefit of the proposed algorithm. MNFBLMS algorithm has wide potential application in ANC systems.