A computational model for MOS prediction

Abstract

A computational model to predict MOS (mean opinion score) of processed speech is proposed. The system measures the distortion of processed speech (compared to the source speech) using a peripheral model of the mammalian auditory system and a psychophysically-inspired measure, and maps the distortion value onto the MOS scale. This paper describes our attempt to derive a "universal", database-independent, distortion-to-MOS mapping function. Preliminary experimental evaluation shows that the performance of the proposed system is comparable with ITU-T recommendation P.861 for clean speech sources, and outperforms the P.861 recommendation for speech sources corrupted by either car or babble noise at 30 dB SNR.