Zizhi Qiao


Voice over Internet Protocol (WIP) applications are becoming more and more popular in the telecommunication market. Packet switched V61P systems have many technical advantages over conventional Public Switched Telephone Network (PSTN), including its efficient and flexible use of the bandwidth, lower cost and enhanced security. However, due to the IP network's "Best Effort" nature, voice quality are not naturally guaranteed in the VoIP services. In fact, most current Vol]P services can not provide as good a voice quality as PSTN. IP Network impairments such as packet loss, delay and jitter affect perceived speech quality as do application layer impairment factors, such as codec rate and audio features. Current perceived Quality of Service (QoS) methods are mainly designed to be used in a PSTN/TDM environment and their performance in V6IP environment is unknown. It is a challenge to measure perceived speech quality correctly in V61P system and to enhance user perceived speech quality for VoIP system. The main goal of this project is to evaluate the accuracy of the existing ITU-T speech quality measurement method (Perceptual Evaluation of Speech Quality - PESQ) in mobile wireless systems in the context of V61P, and to develop novel and efficient methods to enhance the user perceived speech quality for emerging V61P services especially in mobile V61P environment. The main contributions of the thesis are threefold: (1) A new discovery of PESQ errors in mobile VoIP environment. A detailed investigation of PESQ performance in mobile VoIP environment was undertaken and included setting up a PESQ performance evaluation platform and testing over 1800 mobile-to-mobile and mobileto- PSTN calls over a period of three months. The accuracy issues of PESQ algorithm was investigated and main problems causing inaccurate PESQ score (improper time-alignment in the PESQ algorithm) were discovered . Calibration issues for a safe and proper PESQ testing in mobile environment were also discussed in the thesis. (2) A new, simple-to-use, V611Pjit ter buffer algorithm. This was developed and implemented in a commercial mobile handset. The algorithm, called "Play Late Algorithm", adaptively alters the playout delay inside a speech talkspurt without introducing unnecessary extra end-to-end delay. It can be used as a front-end to conventional static or adaptive jitter buffer algorithms to provide improved performance. Results show that the proposed algorithm can increase user perceived quality without consuming too much processing power when tested in live wireless VbIP networks. (3) A new QoS enhancement scheme. The new scheme combines the strengths of adaptive codec bit rate (i. e. AMR 8-modes bit rate) and speech priority marking (i. e. giving high priority for the beginning of a voiced segment). The results gathered on a simulation and emulation test platform shows that the combined method provides a better user perceived speech quality than separate adaptive sender bit rate or packet priority marking methods.

Document Type


Publication Date