Speech Quality Prediction in VoIP Concatenating Multiple Markov-Based Channels

Abstract

The cross interaction of technologies has become standard practice to add more constraints to Voice over Internet Protocol sessions (VoIP). In this paper, a self developed end to end VoIP simulator to predict speech quality is presented. This system has been validated with a real test bench and can provide up to three audio compressors and two channel status at a time. Simulator response is tested through a Mean Opinion Score (MOS) and compared to ITU-T´s G.107 E-model speech quality predictor. High packet loss rate limitation on the E-model is solved by our proposed new parameter. In addition, a methodology to extend this predictor for multiple concatenated channels has been tested and proved to be successful, results in double channels tests range between 1.86R and 8.35R error rate. The methodology is a useful speech quality predictor for design and management purposes, VoIP gateways can take advantage of channels and codec information to guarantee an specific quality of service to end users.