Efficient adaptive bilinear filters for nonlinear active noise control

This paper proposes a novel adaptive bilinear filter-error least mean square (LMS) algorithm and channel-reduced diagonal bilinear filtered-error LMS algorithm, which selectively choose Bilinear channels for coefficient updates in order to reduce computational complexity while still maintaining the performance for nonlinear active noise control. The developed algorithms employ a simple alternative to previously algorithms, which use delays in updating the adaptive filter coefficients and reduce the channels in the diagonal structure. Our experimental results show that both developed bilinear filtered-error least mean square (BFELMS) and channel-reduced diagonal bilinear filtered-error LMS (CRDBFELMS) algorithms gain almost the same performance as compared to diagonal bilinear filtered-x LMS (DBFXLMS) algorithm. What's more, both proposed algorithms could significantly reduce the computational complexity of the standard DBFXLMS algorithms with almost the same performance.