A subspace-based adaptive approach for multichannel equalization of room acoustics

Multichannel equalization (MCEQ) techniques have been proposed for the inversion of acoustic impulse responses (AIRs) with a wide range of applications such as speech dereverberation. Adaptive MCEQ techniques can address the problem of high computational complexity required for the inversion of high order AIRs using non-adaptive algorithms based on the multiple-input/output inverse theorem (MINT). In this paper, we propose a subspace-based adaptive algorithm for MCEQ in the time domain. Unlike conventional techniques, instead of using a single set of inverse filters, multiple sets of inverse filters, each designated for each subspace, are adaptively estimated in our algorithm and equalization is achieved using these filters. Simulation and experimental results show that the proposed adaptive algorithm not only achieves a high convergence rate but also yields better equalization in terms of quality. In addition, the proposed algorithm achieves higher robustness to AIR estimation errors.