Distributed Source Coding System Design Based On LDPC Code

With the development of network and information technology, the traditional joint coding algorithm hasn’t met the needs of many emerging applications, so distributed source coding begins to be concerned about. Distributed source coding as a remarkable focus has enormous research and application value.There are some recent applications such as wireless sensor networks, hyperspectral image compression, single point of view and multi-view video compression. The distributed source coding purpose is to make full use of the correlation between the multiple of source, to the source coding information of maximize compression.DSC requires multiple source independent encoding and joint decoding for all streams using the source correlation at the decoding side to reducing the complexity of the encoding side.The computational complexity will be transferred from the encoding side to the decoding side, to meet the the requirements of the new application.This paper will research source coding in the distributed circumstances. To achieve any point of Slepian-Wolf achievable rate region, according to the the existing LDPC code encoding and decoding algorithm, the paper proposed two rate adaptive LDPC codes based distributed sourcecoding scheme by the principle of coset division.The one is the joint program based on LDPCA code,which uses PEG method to constructe LDPCA code and has the good feature on the loop length and minimum distance.In addition,it can achieve linear encoding while maintaining the same good BER performance. Through simulation experiments show that, using the medium length LDPCA code length has met the performance requirements, and reduced the computational complexity and delay calculation process.The other program is based on QC-LDPC code, which biggest advantage is the check matrix structure with block cycle characteristics and the strong regularity.In the relatively random LDPC code terms, storage about QC-LDPC code parity check matrix H just needed to store the position and the number of cyclic shift for per one sub-block matrix H,which is very helpful to reduce the storage space what the memory occupied, making the description complexity of entire LDPC codes greatly reduce,and bringing great benefits for the design of practical SW encoder.The paper describes the the original logarithmic domain BP coding algorithm, and based on this algorithm, there is proposed a new decoding algorithm. The advantage of this method is to do different operations while updating check node confidence information, in according with the different of coding sequences per-bit information.The simulation shows that the memory space what QC-LDPC code occupied is more small indeed,compared with the traditional LDPC codes,and more more suitable for short-length coding.We can optimize the design of various code rate QC-LDPC code and integrate them into one Slepian-Wolf encoder, without requiring to take up much storage space.Moreover,the paper give the performance simulation in the binary BSC virtual channel for two coding schemes based up LDPC codes.The results show that the solution can achieve the adaptive code rate at the same time that encoding complexity does not change, with good coding performance and more closer to the the Slepian-Wolf bound. For the two schemes, the performance of distributed source coding is better based on QC-LDPC code than LDPCA code. The former take up less storage units with more low energy consumption and more conducive to the applications on energy constrained network.