Adaptive Video Transmission Algorithm Of DASH

With the significant advance of multimedia communication and wireless network-ing techniques, the demand for mobile video is raising up rapidly. Recent researches have demonstrated that video traffic has accounted for 66% of the total mobile traffic. In China, this proportion is up to 69.3%. Traditional video streaming is mainly based on RTP/RTSP protocol. However, it relies on specified streaming servers, which makes it difficult to install, configure and maintain. Meanwhile, the RTP streaming is incom-patible with firewall and NAT. These prevent the wide deployment of RTP/RTSP-based streaming. To overcome these drawbacks, recent years the HTTP-based video stream-ing is becoming a promising technique. Because of interference, user mobility, etc., the condition of wireless channel changes randomly. It is hard to guarantee the smooth-ness of video playback. Supporting video bitrate adaptivity is desired for wireless video streaming. Due to its good compatibility, strong scalability and rapid deployability, Dy-namic Adaptive Streaming over HTTP (DASH) attracts intense research interests from both academia and industry.This thesis will further discuss the rate adaptation for video streaming over wireless networks. The main work can be summarized as follow:1) This thesis has a brief description of the architecture of the video streaming sys-tem. Also, it depicts and compares different video coding standards, formats and transportation protocols which are widely adopted by video streaming. Besides, in this thesis, DASH is introduced in details, including its features and working process.2) This thesis designs and implements the DASH system over wireless networks. This thesis firstly introduces the design principle of DASH, then it describes the design of the system, finally it puts forward the system implement.3) This thesis devises a novel rate adaptation algorithm for DASH over wireless networks. In order to deal with the stochastic fluctuation of channel throughput and video bitrate, it defines the buffer underflow probability. The probability describes the degree of mismatch between channel throughput and video bitrate. Since the occurrence of underflow is expected to be a rare event, based on large deviation principle, the thesis derives an estimation model for underflow proba-bility. Meanwhile, it supposes that the video frame arrival is Poisson distributed which is also widely used in the literatures. The underflow probability can be esti- mated online. Based on the estimated underflow probability, it designs an online rate adaption algorithm.4) In order to verify the performance of the proposed algorithm, this thesis designs and implements a prototype system. Besides this thesis conducts a simulation experiment. Experimental results demonstrate that the algorithm of this thesis can provide a smooth video playback with a high quality.