Estimation of time-variant acoustic feedback paths in in-car communication systems

In-car communication systems aim at improving the speech communication inside a vehicle by picking up the talking passenger´s voice by means of microphones and playing it back over loudspeakers that are located close to the listening passengers. Because of this closed-loop operation, system instability caused by acoustical feedback is a limiting factor for the maximum possible system support. In this contribution, we present a feedback cancellation algorithm based on the NLMS algorithm for estimating the frequency responses of the multiple, time-variant feedback paths. The step-size parameter is calculated time- and frequency dependently and signal decorrelation, which is necessary for convergence of the filters, is performed by a phase modulation filter for frequency shifting. Measurements in a test car prove the effectiveness of the approach in the in-car communication application, even in the presence of background noise.