Perception-based nonlinear loudspeaker compensation through embedded convex optimization

In this paper a novel nonlinear loudspeaker compensation technique is presented which is based on embedded convex optimization. The aim is to compensate for the linear as well as for the nonlinear perceptible distortions incurred in the loudspeaker. To this end, a psychoacoustic model is adopted and a convex optimization based problem formulation is set up. In order to solve the resulting convex optimization problems in a fast and reliable way, a projected gradient optimization method is proposed. From comparative objective evaluation experiments, it is concluded that the proposed nonlinear loudspeaker compensation technique indeed improves the average audio quality scores.