QoS Improvement Schemes for Real-Time Wireless Internet Telephony

Student: Mingkuan Liu

There is a tremendous demand on real-time multimedia delivery over wireless Internet due to the dramatic increase in wireless communication and the growth of the Internet. However, real-time multimedia over wireless Internet poses many challenges. First of all, the inherent best-effort characteristics of packet-switched networks make it difficult to provide guaranteed Quality of Services (QoS) for real-time multimedia delivery. Secondly, wireless channels have much higher packet-loss rate, bit-error rate, and channel instability compared with wired channels due to the noise, multi-path, shadowing fades, user mobility, etc., which results in fluctuating communication channel statistics. Thirdly, real-time communication demands very strict time limitations on the network end-to-end delay and the delay jitter.

The image shown below is a typical Internet telephony system structure. The signaling flow via SIP or H.323 protocol suite is used to set up and control the logical connection between the speaker and receiver. The real-time voice packets are sampled, processed and then transported through the RTP/UDP/IP channel to the receiver side. The receiver does the reverse procedure to recover and then play back the voice.

In this research, we propose an intelligent application architecture and several QoS improvement mechanisms to timely estimate and predict the wireless network traffic statistics and dynamically take appropriate actions to improve the overall performance of a real-time wireless Internet telephony system. The proposed intelligent application framework will actively evaluate the current network situation, collect necessary statistics and give some predictions based on time-series analysis. Based on this real-time updated information, the application's sender side can take some adaptive actions such as voice codec selection, forward error-correction schemes for packet-loss concealment, or header compression to improve the QoS under currently available network resources. Also, the application's receiver side will use these updating traffic statistics to dynamically adjust receiver jitter buffer size, the voice play-out time and the local packet-loss concealment algorithms.

This work is a collaborative effort with Dr. Kevin McNeill, BAE Systems, Inc.