| It is increasingly important to model the VoIP speech quality for technical and commercial reasons.Voice quality can be measured using either subjective or objective methods. Subjective measurement (e.g. MOS) is the benchmark for objective methods, but it is slow, time consuming and expensive. Objective measurement can be intrusive or non-intrusive. As known, many factors affect the quality of audio in IP telephony. We divided these factors in two types:network factors (e.g. packet loss, burstiness, jitter and delay) and non-network impairments (e.g. codec type, echo, noise and gender/language).The main goal of the project is to analyses different factors and develop novel and efficient models for speech quality prediction. For this reason, a new simulation system is set up to collect data. NIST Net is used for network simulation, and OpenPhone simulates communication protocol. This simulation system is more "real'than other researcher's system. The main contributions of the thesis are threefold:1) New Extended E-Models are proposed. One based on standard deviation of packet delay, the other based on language factors. The effect of jitter standard deviation is analyzed, and we find it is accurate by using some function of the standard deviation of the packet delay to describe the relation of delay and voice quality. We also investigate the effects of different language on speech quality in voice over Internet protocol (VoIP) scenarios.There is still a slight difference between diffrent language. A new parameter named IL relates to language is added to E-Model. It is widely accepted the network was self-similar. The degree of self-similarity can be expressed by the Hurst parameter H (0.5 |
PDF Full Text Request