An adaptive subspace based multichannel equalization algorithm for room acoustics equalization with sparseness constraint

Multichannel equalization (MCEQ) techniques are employed in applications such as speech dereverberation to mitigate channel effects. In this paper, we propose a subspace-based adaptive MCEQ algorithm with the aim of addressing the problem of high computational complexity involved in the multiple input/output inverse theorem based non-adaptive MCEQ algorithms. In the proposed algorithm, equalization is achieved using multiple sets of inverse filters such that each set of filters corresponds to each subspace and these filters are adaptively estimated using the least-mean-square type algorithm. In order to improve the convergence performance, the sparseness of the estimated Kronecker delta function is used as a constraint to the adaptive algorithm. Simulation and experimental results show that the proposed adaptive algorithms yield better equalization with improved convergence performance.