Uncertainty of Speech Quality Measurement with PESQ on Speech Enhancement Algorithms

摘要

The purpose of speech enhancement algorithms is to improve speech quality, naturalness and intelligibility by eliminating the background noise and improving signal to noise ratio. This paper investigates the measurement uncertainty of the speech quality achievable by speech enhancement algorithms with the method called Perceptual Evaluation of Speech Quality (PESQ). To this aim, the objective measurement of end-to-end speech quality by means of the PESQ algorithm is introduced, and its accuracy as an estimator of the subjective Mean Opinion Score (MOS) is summarized. Then, the uncertainty of the PESQ results on different speech enhancement algorithms is analyzed. Through analysis of the results, the performance of several speech enhancement algorithms under the car noise conditions is compared, and the reliability of the results is discussed.