Combined Source-Channel Coding With Turbo Codes

The traditional communication system is based on the separability theory, which source coding and channel coding is operated dividually. Such system can not resolve the inherent contradiction between the source coding and channel coding, which is that it will cut the redundancy bits in source coder and it must increase the redundancy bits in the channel coder. In this thesis, a new code method is presented, which takes the affect of-channel noise in the beginning of designing the quantizer and reconstructor pair. The proposed system takes advantage of the available reliability information (generated by the Turbo decoder) to improve the quality of reconstructed signal. The following algorithms have been proposed and tested in the thesis:1, A high efficiency image compress method, named as MMCVQ(Mean-Match Correlation Vector Quatization), is presented, which not only have lower code rate, but also can be operated in shorter time. To take advantage of the relativity of image, this code method use a efficient order, which is that to code the image block on the diagonal at first, then the image block under the diagonal, then above the diagonal. If the image block can code with the same code word as the border upon the image block, then use 2 bits to denote the image block. Thus reach the aim of coding with lower code rate with shorter time. The communication system will use the method to code the source signal.2, To generate the available reliability information, the principle of Turbo coder and Turbo decoder is presented. Turbo coder is composed of tow or more RSC coder connected through several interleavers. The Turbo decoder output is not bit stream, but the log likelihood ratio of each bit. So these value include the channel information, and we can use the information to improve the quality of reconstructed samples.3, Several channel model is presented to design different quantizer and reconstructor pair, and corresponding arithmetic is presented. The objective ofdesigning the quantizer and reconstructor pair is that on the condition of making the system distortion smallest, joint optimizing the quantizer code book and reconstructor code book. It uses algebraic design algorithm and training sequence design algorithm.Joint source coding and channel coding is a developing field, many problem in this field need to be resolved, especially about mobile communication.