Speech Quality Evaluation Of VoIP Based On EPESQ

Due to wide range of services provided by networks, VoIP (Voice over Internet Protocol) has become one of the most prominent applications. The rapid developments of VoIP led to an increased need for evaluating and optimizing the performance of communication system. One of the most important indexes to evaluate the performance of speech communication system is speech quality. Therefore, architecture of the objective evaluation of speech quality, which is smart, rapid and reliable, has become the focus of recent quality measurement research.First of all, the principle of the ITU-T P.862 PESQ algorithm was described in detail .Meanwhile the advancements and the performances of PESQ model were analyzed.We selected the PESQ as the core algorithm for speech assessment of VoIP. On this condition, a new algorithm which is called a high-performance algorithm, Efficient Psychoacoustics Evaluation of Speech Quality (EPESQ), based on MFCC was studied and improved.Theoretical analysis and experimental results showed that the EPESQ algorithm has comparable performance with PESQ. However, its time complexity and space complexity are much lower than PESQ. Finally, a C/C++ implementation of this algorithm was integrated into an objective evaluation platform of VoIP speech quality. In order to construct a complete architecture of speech quality assessment of real-time online VoIP, a method of effective payload replacement which acted as the interface between VoIP system and EPESQ was proposed. After that, some applications of the architecture were discussed.The experiment results showed that, the speech quality evaluation platform based on EPESQ has good performances, and due to the low complexity of EPESQ, it is suitable for real-time speech quality evaluation and implementation on embedded system. In order to improve the VoIP services, the complete architecture of speech quality assessment must be studied further.