Output-based objective measure for non-intrusive speech quality evaluation

This paper describes a newly developed output-based method for non-intrusive evaluation of speech quality of voice communication systems, and evaluates its performance. The method, which uses only the output of the system, is based on measuring perceptually motivated objective auditory distances between the voiced parts of the speech signal whose quality to be evaluated to appropriately matching reference vectors extracted from a pre-formulated codebook. The codebook is formed by optimally clustering large number of perceptually-based parametric vectors extracted from a database of clean speech signals. The auditory distance measures are then mapped into equivalent subjective score, represented by the Mean Opinion scores (MOS), using regression. The required clustering and matching processes are achieved by using an efficient neural network based data mining technique known as the Self-Organizing Map. Perceptual, speaker-independent parametric representation of the speech is achieved by using Linear Prediction (PLP) and Bark Spectrum analysis. Reported evaluation results show that the proposed system is robust against speaker, utterance and distortion variations, and outperforms the ITU-T P.862 Perceptual Evaluation of Speech Quality (PESQ) for cases of speech degraded by channel impairments.