One of the most challenging problems to address in the modeling of acoustic channels is the presence of nonlinearities generated by loudspeakers. This problem has become even more widespread due to the growing availability of low-cost devices that introduce a larger amount of distortions and decrease the quality of hands-free speech communication. In order to reduce the effect of loudspeaker nonlinearities on the speech quality, nonlinear acoustic echo cancellers are adopted. In this paper we present a new adaptive filtering structure for the reduction of nonlinearities in the acoustic path, based on a nonlinear transformation of the input audio signal by means of functional links. We use such a nonlinear model in conjunction with a linear filter providing a nonlinear adaptive architecture for acoustic applications. Experimental results prove the effectiveness of the proposed model in reducing loudspeaker nonlinearities affecting speech signals.
This paper costs $33 for non-members and is free for AES members and E-Library subscribers.