A modified Itakura speech distortion measure based on auditory properties

Abstract A modified Itakura speech distortion measure is presented in this paper. This new modified Itakura speech distortion (called M-Itakura) has taken into account auditory properties of human beings. In the algorithm of the M-Itakura measure, the speech signal is processed using results of psychoacousitic measurements, which include critical band analysis, equal-loudness pre-emphasis and intensity–loudness power law, and the short-time speech power spectrum is converted to auditory perceptual spectrum. Preliminary experiments in comparison with the Itakura measure have been done. Simulation results have shown an improvement of the M-Itakura measure over the conventional Itakura measure and the M-Itakura measure is more consistent with subjective assessment of speech quality.