The frequency-domain adaptive Kalman filter (FDAKF) is a popular choice for multichannel acoustic echo cancellation (AEC) due to its good initial convergence and robustness to double-talk. However, without additional measures its reconvergence and tracking capabilities are known to be suboptimal. Previous studies have particularly focused on abrupt echo path changes and have proposed different methods to optimize the filter’s reconvergence. Motivated by our application of an acoustic echo cancellation system for immersive telepresence, this paper investigates continuous echo path changes caused by moving microphones. The echo cancellation performance of the FDAKF is studied for different parameters of the underlying model inside the Kalman filter. Experimental results show, that even in the very challenging scenario of a moving microphone, a small echo reduction can still be achieved with suitable parameters for the considered microphone velocities. Furthermore, a novel method is proposed, which includes a microphone motion-controlled online parameter switching for the FDAKF by means of external motion sensors. In this paper the method is studied within a proof-of-concept. Experiments show a behavior matched to static and dynamic phases and even an increased reconvergence speed in the transition from dynamic to more static phases.
https://www.aes.org/e-lib/browse.cfm?elib=21667
Click to purchase paper as a non-member or login as an AES member. If your company or school subscribes to the E-Library then switch to the institutional version. If you are not an AES member and would like to subscribe to the E-Library then Join the AES!
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.
Learn more about the AES E-Library
Start a discussion about this paper!
