High-performance Decoding Algorithm For Polar Codes

Polar code was first proposed by Arikan in 2008.It is proved have capacity-achieving performance.Recent years,many scholars have focused on the hardware implementation of the successive cancellation(SC)algorithm and the list SC(SCL)algorithm for polar codes.How to improve the throughput of the polar decoders becomes a hot topic.Nowadays,many decoders for polar codes are proposed to optimize the architecture of decoders,the throughput of the decoder remains unsatisfied,which is the main purpose of this graduation design.It's not hard to find there are two main ways to improve the decoding throughput after a careful study for polar codes.One is to optimize the processing element of polar decoder and the other one is to optimize the whole decoding process,which is exactly the two main research points of this paper.The first research point is to optimize the internal structure of the decoder using stochastic computation.Stochastic computation transforms the serial operations inside the processing elements into parallel operations,which reduce the delay of each processing element and improves the decoding throughput.In this paper,stochastic computation is extended into the decoding of the polar cods.A family of SC decoders based on stochastic computation is proposed,including decoders based on bipolar stochastic computation,low bits stochastic computation and multi-level stochastic computation,which improves the throughput and resource utilization of polar decoders.Compared with fixed-point decoders,the proposed decoder based on multi-level stochastic computation improves the throughput by 20%while having 0.5dB loss in decoding performance.Another research points of this paper is to combine decoding iterations,and optimize the decoding process of the polar codes using look-ahead algorithm.A large number of log likelihood ratios are generated during the decoding procedure.These values need to be re-calculated thus a large number of registers are needed to store these temporary variables.So the two or more decoding levels can be merged into one,which can greatly reduce the number of registers and decoding cycles.In addition,the throughput of polar codes can be further improved while combined with look-ahead algorithm.Also this proposed mixed-radix decoder is a good tradeoff between hardware resource and throughput,which can meet different design requirements.These two kinds of decoding algorithm are efficient ways to improve the throughput and reduce resource consumption for polar decoders.What's more,they both have good universality and can be inserted into other high-throughput schemes to further improve decoding performance.The simulation and comparison results show these two methods has the prospects for practical applications.