Double-talk robust acoustic echo cancellation for CD-quality hands-free videoconferencing system

A novel monaural acoustic echo cancellation method developed for a 20-kHz wideband hands-free videoor teleconferencing system is presented. The method can effectively reduce undesired acoustic echo included in a signal arriving at a microphone from a loudspeaker and can emphasize the target talker´s voice when near-end and farend talkers speak simultaneously (i.e. double-talk). The method estimates a power frequency response of an acoustic echo path between a loudspeaker and microphone (acoustic coupling level) whether or not double-talk has occurred, then calculates a post-filter that effectively reduces the undesired echo. In the echo cancellation processing, the calculation complexity is reduced to make the processing suitable for real-time implementation by using a low-complexity subband approach that employs a new subband filtering algorithm. Experiments were conducted to examine the performance of the proposed method. The results indicated that the proposed method delivered natural-sounding near-end speech even during double-talk periods while sufficiently suppressing the undesired echo.