| Summary: | 碩士 === 國立臺灣科技大學 === 電機工程系 === 104 === With the prevalence of video streaming services (e.g. YouTube, Netflix), cloud video streaming technologies become important. HTTP-based adaptive streaming (HAS) has emerged as the prominent technology for video delivery over the Internet. The HAS allows video streaming services to go through firewalls and NAT friendly and has the potential to improve the quality of service (QoS), as comparison with the traditional centralized streaming service. In the HAS approach, video contents are segmented into small packets with fixed time duration and then encoded at multiple resolutions and bitrates to meet the requirements of user network conditions and heterogeneous devices. In this thesis, we design and implement a cloud HTTP adaptive streaming system based on MPEG-DASH standard, which includes a server and multiple DASH clients. We also propose a client-side adaptive algorithm, based on network throughput and future buffer usage (NTFB). As multiple clients share the same network resources and compete for available bandwidth, the NTFB will select the required video segment from a best client, which compromises stability, fairness, and efficiency for robust video adaption. Experimental results show that, as compared with conventional algorithms, the NTFB can yield smoother bitrate transmission and higher video PSNRs with a smaller buffer.
|
|---|
