Towards Wave-Domain Adaptive Filtering for Multichannel Acoustic Echo Cancellation in Higher-Order Ambisonics Systems
PubDate: December 2021
Teams: Institute of Communications Technology Leibniz University Hannover
Writers: Marcel Nophut; Robert Hupke; Stephan Preihs; Jürgen Peissig
Abstract
The concept of wave-domain adaptive filtering (WDAF) is a powerful technique in the field of massive-multichannel acoustic system identification. By using fundamental solutions of the wave-equation for signal representation the wave-domain system model exhibits desirable properties in suitable setups. This allows efficient approximations of the model or can improve the convergence of an adaptive filter. The applicability of the WDAF technique to higher-order Ambisonics systems and three-dimensional wave fields was often mentioned in literature, but conducted experiments were always limited to two-dimensional setups. This paper investigates the WDAF technique in a new context of practical Ambisonics systems, aiming at acoustic echo cancellation for immersive telepresence systems. Experiments are conducted with simulated and measured room impulse responses in ideal and non-ideal loudspeaker setups. Furthermore different Ambisonics decoding methods are investigated and compared. The transferability of previously proposed WDAF approaches is examined and the results show the general applicability of these approaches, but with some limitations for real-world problems.