Non-Intrusive Quality Evaluation of VoIP Using Genetic Programming

Packet loss is bursty in nature and it affects the quality of VoIP adversely. In this article a novel approach based on genetic programming (GP) has been presented that maps the effect of bursty packet loss on listeners´ perception of speech quality. The ITU-T P.862 (PESQ) algorithm is used as a reference model in this research. A simplistic non-linear model for effective VoIP quality estimation has been proposed. The performance of the model shows a high correlation with the PESQ algorithm. In the light of current and related research it has also been shown that PESQ does not model the effect of bursty packet loss on speech quality. However, as a good approximation to PESQ the proposed model is suitable for reliable, real-time and continuous evaluation of VoIP calls. It is also claimed that it is conveniently deployable on a wide variety of hardware platforms