Acoustic echo cancellation using blind source separation principles.

摘要

The goal of this project was to develop acoustic echo cancellation (AEC) algorithms based on the blind source separation (BSS) principle. Traditional AEC systems are based on adaptive filters and double talk detectors. A double talk detector plays an essential role in an acoustic echo cancellation system. It helps to maintain the system's stability when a double talk situation is present. Although many efficient, fast convergent adaptive filters are available, no reliable talk detector has ever been proposed. With the blind source separation principle, we are able to remove the double talk detector from the acoustic echo cancellation application. Thus, we are able to obtain the convergent algorithm without worrying about the double talk situation.