Integrating acoustic echo cancellation with adaptive beamforming microphone arrays
Walter Kellermann · The Journal of the Acoustical Society of America · 1999
In hands-free and untethered man–machine speed communication, acoustic echo control (AEC) is widely acknowledged as a critical issue determining, e.g., the quality of a full-duplex dialogue, or the speech recognition performance of voice control applications. In comparison to single microphones, microphone arrays (MAs) allow for an improved acoustic front-end by increased spatial selectivity and the flexibility to track moving sources. It is argued that in most cases, however, this will not suffice to eliminate the need for AEC. The most effective solution to AEC is provided by echo cancellation systems, which model the acoustic echo path from the source to the receiver by an adaptive digital filter. While the echo path for a time-invariant MA can be modeled by a single adaptive filter, adaptive beamforming MAs require a time-variant filter capable of tracking the change of the beamforming characteristic instantaneously or, alternatively, one echo canceller for each sensor of the array. Given the considerable computational load for acoustic echo cancelers, a concept is proposed here that limits the beamforming to several fixed beams and foresees one echo canceler for each beam. In addition to computational efficiency, this system provides synergies enhancing the control of the acoustic echo canceler as well as source finding for beamforming.