Performance of auditory features using MFCC and Teager-Kaiser energy computation for non-intrusive speech quality assessment

This paper aims to bring in light non-intrusive speech quality assessment using Teager-Kaiser energy computation. Based on the above mentioned energy computation technique, Features in the form of cepstral coefficient are calculated and thereby, compared to the classical Mel-frequency cepstral co-efficient. The energy computation technique is vividly used in automatic speech recognition area. The Gaussian Mixture Model is used for finding the respective Features and hence correlation between the subjective and objective Mean Opinion Score (MOS) is found. Its performance is evaluated on ITU-T Supplement-23 database.