J.M. Wesselink

Fast affine projections and the regularized modified filtered-error algorithm in multichannel active noise control

In this paper, real-time results are given for broadband multichannel active noise control using the regularized modified filtered-error algorithm. As compared to the standard filtered-error algorithm, the improved convergence rate and stability of the algorithm are obtained by using an inner-outer factorization of the transfer path between the actuators and the error sensors, combined with a delay compensation technique using double control filters and a regularization technique that preserves the factorization properties. The latter techniques allow the use of relatively simple and efficient adaptation schemes in which filtering of the reference signals is unnecessary. Results are given for a multichannel adaptive feedback implementation based on the internal model control principle. In feedforward systems based on this algorithm, colored reference signals may lead to reduced convergence rates. An adaptive extension based on the use of affine projections is presented, for which real-time results and simulations are given, showing the improved convergence rates of the regularized modified filtered-error algorithm for colored reference signals.

Rapidly converging adaptive state‐space‐based multichannel active noise control algorithm for reduction of broadband noise

Rapidly changing spectra may lead to performance limitations in adaptive systems for broadband active noise control, especially in multichannel systems. This paper presents techniques to address the negative consequences of two main causes. First, the dynamics of the transfer paths between the noise control sources and the error microphones is compensated for by using a regularized state‐space based adaptive filtered‐error scheme. Second, for the reference signals a modified adaptive scheme is used, taking into account the nonwhiteness of these signals as well as the correlation between the individual signals. Examples are given for simulated data and for real‐time implementations. [This work was partly supported by EC Contract 501084 (InMAR).]