Enhanced Itakura measure incorporating masking properties of human auditory system

A new enhanced Itakura (E-Itakura) speech distortion measure is proposed in this paper. It incorporates masking properties of the human auditory system into the original Itakura measure. Inaudible noise components masked by speech signals are excluded from the calculation of the E-Itakura measure, while the intrinsic advantage of the Itakura measure is retained. The proposed new measure has been compared with the original Itakura distortion, frequency-weighted Itakura spectral distortion, cepstral distance and Bark spectral distortion measures. The comparison results show that the correlation between the original Itakura measure with speech quality has been improved from 0.73 to 0.89 with the incorporation of the enhancement feature, and that the E-Itakura measure offers a more consistent indication of the subjective quality of speech. (C) 2003 Elsevier Science B.V. All rights reserved.