Abstract
With the tremendous growth in the volume of video traffic in mobile networks with restricted network bandwidth and fluctuating channel quality, it becomes a critical issue to effectively adapt video streaming in wireless networks to support better video streaming performance. Dynamic Adaptive Streaming over HTTP (DASH) is the most common solution to adapt video bit rate according to available bandwidth to improve video performance. However, DASH selects video versions only based on the average bit rate of a whole video, which is a coarse-grained streaming scheme, instead of exploiting the variable bit rates of a video. Thus, DASH cannot adapt video streaming properly enough to maintain video quality in unstable channel environments. In this thesis, we propose a Profile-based Adaptive Video Streaming (PAVS) scheme to effectively adapt video streaming in wireless networks. Compared to DASH, PAVS retrieves the average bit rate of each video segment to get fine-grained video information. Moreover, PAVS estimates the expected bandwidth in wireless networks by tracking TCP ACKs from clients and using the historical network bandwidth information between a video server and clients; then, it selects the proper video version according to the fine-grained video information, the expected bandwidth, and buffer occupancy of the client. In summary, the PAVS provides a fine-grained video version selection and minimizes control message in resource restricted wireless networks. Compared to DASH, based on our experimented results, PAVS reduces the number of video re-buffering events by 76% and the total time of video re-buffering by 81%. Besides, PAVS increases the bandwidth utilization of adapting video streaming to 84%.