Convergence rate of stereo acoustic echo cancelers with pre-processing

摘要

The convergence properties of stereo acoustic echo cancelers with pre-processing are examined. In stereo acoustic echo cancelers, the non-uniqueness problem occurs since there are two paths from the talker to a speaker. Some methods have been proposed to cope with this based on time-varying pre-processing of the input signals. Since pre-processing degrades the quality of input signals, we need to find out a method with small degradation and high increase of convergence speed. The input-sliding method is such one but its mechanism is unknown. We derive the theoretical convergence rate of stereo echo cancelers with pre-processing and confirm the validity by computer simulations.