Article details

Research area
Speech enhancement

Jahrestagung für Akustik der Deutschen Gesellschaft für Akustik


Timo Matheja, Markus Buck, Tobias Wolff

Robust adaptive cancellation of interfering speakers for distributed microphone systems in cars


Hands-free systems in cars aim to capture the speech of different speakers at the best. Therefor distributed microphones can be aligned to each of these speakers and can be mounted in their vicinity. One application of this setup is to optimize speech for speech recognition. It is desirable to cancel the influence of possible interfering speakers in the microphone signals. For this task an adaptive filter is used to cancel the interfering signal from the target one. In contrast to the very similar and well known echo cancellation problem a noise component also occurs on the filter input signal. In this paper an optimal step size is proposed that considers this noise component for controlling the adaptation of a normalized least mean square (NLMS) algorithm in the short-time frequency domain. A Signal-to-Interference-Ratio (SIR) based adaptation control is also investigated. The performance of both approaches is evaluated.

Read/download now