Acoustic Howling Detection and Suppression for IP-Based Teleconference Systems

Abstract:
IP-based teleconferencing scenarios with multiple participating devices in the same room like in an open-plan office cause a new type of feedback problem. Due to the high system delay over the network, conventional feedback suppression techniques cannot be used. Therefore, in this paper a new method for detection and suppression of acoustic howling caused by feedback is proposed. For detection a combination of correlation- and power-based features is utilized, controlling adaptive notch filters for the actual suppression. It is shown how the detection performs on real world signals. The full system is applied in a simulation framework, showing the effectiveness for howling detection and suppression.