The uncertainty entropy of low-rate speech quality evaluation and the analyses of the gray correlation

Abstract

The low-rate speech quality evaluation has an obvious error for the uncertain factors with multi-source, heterogeneous and time-varying. This paper proposed a new model to measure the error of the speech quality evaluation based on entropy (SQEE). The gray correlation analysis among the factors but also the uncertain entropy was designed to keep the model optimization and equivalence. What’s more, the parameters of determinacy and sensitivity were proposed to measure the accuracy and the efficiency. After the theory and simulation, the new method enhanced the error of the traditional methods (PESQ, PQSM) about 30% below the 250 ms-delay.