Research On The Theory And Applications Of Distributed Source Coding

With the development of some new technologies, such as computers, communications, networks and multimedia, how to effectively compress data has became an important part of digital communications and attracted more and more attentions. What is noticeable, in the communication system of 21th century, the technical support to multimedia which includes audio and video, will become an inevitable trend. However, there are two problems in the realization of one practical system:first, it needs to compress the source with high efficiency so as to utilize the limited channel band, called source coding problem. Second, protection is necessary for the compressed data to resist the loss or damage brought by channel error. This is called a channel coding problem. The enhancement of. compression rate of source coding will consequentially decrease its capability of error counteraction. On the contrary, the increasing of error resist capability will be at the cost of compression efficiency. In fact, the spirit of source coding and channel coding, in some sense, is contradictive. Base on the research of the theory of source coding and channel coding, applying wireless sensor networks as the application scenario, this thesis concentrates on the problem of distributed source coding, that is, containing the same coding effect as traditional source coding, the strategy of increasing compression rate and decreasing the complexity of encoder. In addition, the joint source and channel coding problem in wireless channel is considered and extended, combined with distributed source coding theory, to the research of distributed joint source and channel coding.The major work in this thesis is as follows:1. Research on distributed source coding based on general coding theory is addressed. According to the traits of distributed source coding theory, a decoding algorithm based on universal coding theory is proposed along with the analysis on decoding complexity and comparison with traditional method.2. An unequal error protection scheme based on distributed source coding is proposed in consideration of the traits of wireless sensor networks, that is, large scale, low computational capability and limited energy. Compared with conventional unequal error protection schemes, this scheme just need to design one pair of encoder and decoder for multiple sources so as to extremely reduce the design complexity of large scale wireless sensor networks. What is more, it also owns the property of rate adaptive.3. Considering that the unequal error protection scheme mentioned above is based on asymmetric distributed source coding, we extend the coding approach to the case of symmetric distributed source coding and propose a new architecture of unequal error protection in wireless sensor networks. The architecture is then generalized to the case of joint coding of multiple sources. Simulation results show that, besides the same decoding performance and unequal error protection capability, the symmetric coding based scheme is more practical for its flexible rate allocation property.4. Make improvement on the traditional joint source and channel coding scheme and put forward a double-level error resilient joint source and channel coding scheme for image transmission in wireless channels. The improvement nearly eliminates the problem of loss of data boundary. A better image recovery effect is achieved by inserting one module between source coding and channel coding to re-structure the data transmitted and result in the self-synchronization at decoder side.5. Integrate the theory of distributed source coding and joint source and channel coding into distributed joint source and channel coding. Based on the research of the method of syndrome, this thesis propose a syndrome-based distributed joint source and channel coding scheme, other than traditional used parity-based method. Simulation results illustrate the efficiency of the new scheme not only in the aspect of same decoding performance as parity-based scheme, but also due to its coding optimality.6. Investigate the problem of correlation estimation in distributed source coding, including the construction of theorical model and analysis of the two cases of correlation estimation. Specially, emphasis is put on the analysis of influence to decoding performance caused by the case of under estimation.