Abstract
In multichannel acoustic echo canceler based on linear combination, the uniqueness problem exists that adaptive filter's can't identify the correct echo-paths due to highly cross-correlated multichannel input signals. Pre-processing for input signal are good candidates for solving this problem. However, it can have the drawbacks to deteriorate audio quality as a secondary problem. In this mis-adjustment problem, the universal solution is not led in the conventional adaptive filter's optimization. In this paper, we propose a new multichannel adaptive echo canceling algorithm based on particle swarm optimization (PSO) without pre-processing to input signals. PSO is a stochastic optimization technique inspired by social behavior of bird frocking or fish schooling, and this is the first attempt to apply it for multichannel acoustic echo cancelers. By performing various simulation we confirm the proposed method can be estimate correct echo-paths with highly cross-correlated input signals.
