Enhanced PESQ algorithm for objective assessment of speech quality at a continuous varying delay

Abstract

ITU-T P.862 - “Perceptual Evaluation of Speech Quality (PESQ)” is well known as an intrusive objective speech quality assessment method. Some reports have found that the PESQ time alignment mechanism fails to estimate delay where signals with high packet loss rate and dynamic time processing are present. A new time-alignment algorithm to improve the PESQ accuracy for time-scale modified voice transmission is suggested here. In the propose model, the time alignment of reference and degraded speech is estimated using Dynamic Time- Warping (DTW) in contrast to correlation and splitting methods used in the standard PESQ. Comparative results versus subjective Mean Opinion Score (MOS) show improvement in cases where dynamic time processing of signals is present.