Combination of filtered-x adaptive filters for nonlinear listening-room compensation

Abstract

Audio quality in sound reproduction systems can be severely degraded due to system nonlinearities and reverberation effects. In this context, linearization of loudspeakers has been deeply investigated but its combination with room equalization is not straightforward, mainly when the nonlinearities present memory. In this paper, a method relying on the convex combination of two linear filters using the filtered-x LMS (FXLMS) algorithm and based on the virtual path concept to preprocess audio signals is presented for nonlinear room compensation. It is shown that the combination of two linear adaptive filters behaves similarly to the filtered-x second-order adaptive Volterra (NFXLMS) filter. Moreover the new approach is computationally more efficient and avoids the generation of higher harmonics. Experimental results validate the performance of the new approach.