Research On Polar Code Decoding Based On Stochastic Computing

Polar codes have the best theoretical performance,lower decoding complexity and other advantages compared to other channel coding methods.Polar codes have been significantly attended in the past few years.Polar codes can be decoded using the successive cancellation or belief propagation decoding method.However,the successive cancellation decoding method suffers from long latency due to the sequential decoding scheme.The belief propagation algorithm based on the probabilistic reasoning is one of the polar codes decoding methods.Polar codes can be treated-as a coding method based on the graph model.Therefore,polar codes can be decoded using the belief propagation method.And the belief propagation decoders can reduce the decoding delay with its parallel decoding scheme.In this thesis,we study the belief propagation decoding method based on stochastic computing for the polar codes.Firstly,the history and background knowledge of channel codes are reviewed.We also introduce the research statues of the polar codes.Polar codes,which are the first provable capacity-achieving channel codes,are based on phenomenon of channel polarization.Then the successive cancellation algorithm with simple structure and low complexity is introduced.The belief propagation algorithm based on graph model is also introduced.Compared to the successive cancellation algorithm,the belief propagation decoders can be performed in the a-parallel manner to achieve short latency and high throughput.However,the belief propagation decoding algorithm suffers from high implementation complexity.The stochastic computing can mitigate this challenge with simple stochastic computing circuits.The simulation results show that the optimized belief propagation decoders with stochastic computing can achieve similar performance with the original scheme.Finally,we ues the high accuracy stochastic computing method based on segmentation method solve the problem of long sequences in traditional stochastic computing.The simulation results show that this scheme can reduce the length of random sequences with the guarantee of decoding performance.