Quality Improvement of Mobile Video Using Geo-Intelligent Rate Adaptation

Abstract

Adaptive video is a popular technique to continuously deliver a video stream to a user in the best quality possible when the underlying network bandwidth cannot be guaranteed. As such, quality of adaptive video depends critically on the agility of the rate adaptation algorithms in tracking the varying bandwidth. In this paper, we investigate the performance of a popular rate adaptation algorithm, namely, TCP-friendly rate control (TFRC), in vehicular environments. Our results show that TFRC cannot cope well with the pattern of bandwidth changes faced by a user travelling in a fast moving vehicle, resulting in poor viewing experience. Motivated by the observation that bandwidth changes in vehicular environment is significantly influenced by the rapid change of user's geographic location, we propose Geo-TFRC, which empowers TFRC with a street-level bandwidth map that holds summary of past bandwidth observations for each segment of the street. We conduct simulation experiments which are driven by the real High-Speed Downlink Packet Access (HSDPA) bandwidth traces collected from a vehicle traveling along a route in Sydney. Our results reveal that Geo-TFRC can track the bandwidth changes much more effectively, which in turn improves the quality of the mobile video. We find our proactive approach can significantly reduce the time that a user suffers from pixelated viewing experience by up to five folds as compared to TFRC.