Research On Codec Technology Of Fountain Codes In Erasure Channel

Fountain codes is an important class of Erasure Codes, and has drawn increasing attention and research efforts in the academia and industry due to its non-fixed rate, linear encoding and decoding complexity, and no feedback requirements, etc. Raptor codes is a most important type of fountain codes, in this dissertation, we focus on the study of Raptor codec technology in erasure channel for applications in MBMS and relay system. In particular, we designed improved decoding algorithm, degree distribution of relay system, Unequal Error Protection (UEP) Raptor codes, and gave the simulation and evaluation for performance of several types of application scenarios. Our research made certain contribution to promote the practical applications of Raptor codes. The main results of this work are as follows:Firstly, we propose an improved decoding algorithm, namely Y algorithm, after in-depth study of the current decoding algorithm of Raptor codes, which can effectively balance the computational complexity of decoding and the decoding failure rate. In the system with a simple feedback channel, we present a HARQ_Raptor solution, which improves the decoding success rate through NACK, and thus enhances the decoding performance.Secondly, we establish a C-based simulation platform for assessing the performance of Raptor codes. We have carried out lots of simulation for Raptor codes'codec algorithms and various types of application scenarios, and gained several useful conclusions for Raptor codes. The simulation program can be easily transplanted into practical systems as a function module in which Raptor codes can be used as a FEC.Thirdly, we propose a degree distribution optimized algorithm for Raptor codes based on the three nodes relay system. After a series of theoretical reasoning and proofs, the algorithm can be summarized as an optimization problem. Simulation results show that the improved degree distribution improved Raptor codes'decoding performance significantly.Fourthly, we propose a method about how to choose parameters which is very important for UEP-Raptor codes'designing, and thus give a new UEP-Raptor codes'degree distribution designing algorithm. The algorithm can be summarized as solving an optimization problem, which can be simplely solved by matlab and other mathematical tools. Simulation results show that the UEP-Raptor codes designed using this algorithm outperformance the existing UEP-Raptor decoding methods.Finally, we propose an UEP-Raptor codes'optimization algorithm based on the image transmission. Simulation results show that, optimized UEP-Raptor codes designed using this algorithm can effectively improves the clarity of the image in receiving end, so the optimized UEP-Raptor codes is superior to the general UEP-Raptor codes. In the final chapter of this dissertation, we analyzed the limitations of existing research, and prospected the research directions and the issues to be resolved in the future.