An Improved Selective Active Noise Control Algorithm Based on Empirical Wavelet Transform

The gradual adaptation and possibility of divergence have been the two main obstacles in the efficient implementation of conventional adaptive active noise control (ANC) to a wider range of applications. Selective ANC (SANC) has been proposed to rapidly reduce noise by selecting a pre-trained control filter for different primary noise detected, and improve the robustness of the system. For stationary noise, considerable noise reduction performance and system stability are obtained by SANC. However, for non-stationary noise, in order to track the variability of the signal, frequency-band-match and selection have to be conducted constantly, which results in high computational burden. To confront this problem, empirical wavelet transform (EWT) is introduced to simplify the matching and selection of SANC in this paper. This EWT based SANC (SANC_EWT) algorithm extracts the first mode of random noises, and attenuates the noise immediately by picking the optimal pre-trained control filter labeled by the first boundary. Therefore, computational complexity is reduced drastically. Simulation results show that convergence could be reached rapidly. Better noise reduction performance is achieved by SANC_EWT compared to conventional FxLMS and SANC algorithms.