Font Size: a A A

Research On Multi-Core Parallel Computing In Streaming Media Service System

Posted on:2010-08-21Degree:MasterType:Thesis
Country:ChinaCandidate:C H WeiFull Text:PDF
GTID:2178360278973869Subject:Computer system architecture
Abstract/Summary:PDF Full Text Request
Now people are in high demands of streaming media technology, so how to improve the real-time video system is the all-important task. The way to solve this problem, on the one hand is to enhance network bandwidth, on the other hand is to raise the compression rate and to optimize the transport protocols. At present, many chip manufacturers at home and abroad are committed to the development of multi-core technology. Multi-core chips can increase the frequency of this system and reduce the power consumption at the same time, increase the speed of computing. In addition, parallel computing technology is being developed quickly these years. Parallel computing technology divides the computing tasks into pieces reasonably, so the computing resources can be utilized sufficiently and the complex computing tasks can be resolved efficiently. In order to improve the speed of compression, in conjunction with the current rapid development of multi-core technology for the compression process to be improved.First, this thesis presents the architecture of streaming media service system, including obtaining the video streaming, video encoding and the transport protocol between the server and client. This thesis mainly researches on the last two parts. At the same time, this thesis analyses MPEG video compression algorithms and motion compensation prediction.Then, through the study on the MPEG-4 standard compression algorithm, we find that there are much parallel in many compression algorithms. Parallel is mainly reflected in two aspects. One is the separate coded of the video object in the encoder, the other is the motion estimation algorithm among the VOP. For the first aspect, this thesis proposed a video object schedule algorithm to ensure the synchronization of the video objects during the parallel encode process; while for the second aspect, this thesis improved the serial directional diamond motion estimation algorithm, got the directional diamond motion estimation algorithm based on parallelism and prediction (PPDDME), then simulated this algorithm on the double-core platform UP-OMAP5910 and on the PC using multi threads. The M&D and Stefan frames are used in the experiment. The whole frame is taken as a video object plane. The experimental results show that, after the using of parallel algorithm, compression speed is about 3 times faster than the serial directional diamond motion estimation algorithm. Considering the communication among the threads, the improvement is distinct.Finally, this thesis analyzes the transport protocols used in the streaming media service, and take the RTP/RTCP as the protocol of this system. In the application, RTP is been used in the data transport between the server and the client, while RTCP is been used in the control information transport. RTP get the packet loss rate, network jitter and the receive ability of the client through an adaptive algorithm. The receive ability of the client includes the packet sort efficiency and the decode efficiency. This thesis takes the three factors into account, gets an integrated network state scale algorithm. This algorithm considers both the long-time network state and the current network state, as well as the receive ability of the client. Then RTCP send the information to the server. The server adjusts transmitting rate according to these information. The thesis gives a detailed analysis about the modules function of the protocol.
Keywords/Search Tags:MPEG-4 Standard, Multi-core Processor, Parallel Computing, Motion Estimation, RTP/RTCP Protocol
PDF Full Text Request
Related items