A Simplified Subband ANC Algorithm Without Secondary Path Modeling

Min Gao,Jing Lu,Xiaojun Qiu

DOI: https://doi.org/10.1109/taslp.2016.2516439

2016-01-01

Abstract:Active noise control algorithms without secondary path modeling are very appealing because they can effectively overcome the problems caused by imperfect secondary path model. For broadband noise control, subband structures have been adopted in these kinds of algorithms, where the full-band adaptive filter coefficients are obtained from subband filter coefficients with weight transformation. However, the usually used two-direction search method suffers from slow convergence if the secondary path phase is close to ±90°. To deal with this problem, a four-direction search method has been proposed in frequency domain, which is sometimes too complicated for practical applications. In this paper, a simplified algorithm is proposed to reduce the implementation complexity. In the proposed algorithm, the delay of the secondary path is estimated first, and then the estimated delay is used in the delay compensation to reduce the number of subbands. Two reference signals are generated in each subband and two update directions are used. Only one subband reference signal and one update direction are selected to approximate the phase response of the residual secondary path, and then the full-band adaptive controller coefficients are adjusted directly in time domain with them without the need of further weight transformation. Just like those existing subband algorithms, the simplified algorithm can obtain similar steady-state noise reduction performance as that of the FxLMS algorithm but without secondary path modeling. Simulations and experiments are carried out to demonstrate the feasibility of the proposed algorithm.