Loudspeaker linearization using perceptual distortion measures

Volterra filters are used to model loudspeaker nonlinearities. To compensate for the nonlinearities of the electro-acoustic conversion, the signal is passed through a predistortion filter placed between the audio signal source and the loudspeaker. Predistortion filters based on the inverse of second order Volterra filter increase the signal to noise ratio but do not significantly enhance the perceptual quality. This paper presents predistortion filters based on the cepstral and symmetrized Itakura-Saito spectral distortion measures. It is shown that the proposed predistortion filters enhance the subjective quality. Improvements of over 0.5 MOS can be achieved for a second-order memoryless nonlinearity.