Study Of Cascaded Decoding Algorithms For Polar Codes Based On Belief Propagation

Polar codes as an important technology for next-generation mobile communication systems,has been selected as the control channel coding standard for the 5G Enhanced Mobile Broadband（e MBB）scenario.The advantage of polar codes is that in addition to the reachable channel capacity,there is no error floor in serial cancellation（SC）decoding.However,the serial nature of SC decoding makes the SC algorithm and its variants suffer from a large decoding delay.The Belief Propagation（BP）decoding algorithm has excellent performance in terms of latency due to high parallelism,but suffers from the problem of insufficient decoding performance.In this thesis,the following research is conducted based on the traditional BP decoding algorithm.1.An improved algorithm is proposed to cascade it with Ordered Statistic Decoding（OSD）.The algorithm reduces the oscillation of information by optimizing the information propagation between the two cascade algorithms,thus improving the overall performance.Simulation results show a 0.2 d B performance improvement over the original BP-OSD at a Freak Error Rate（FER）of10^-3.2.A segmented Cyclic Redundancy Check（CRC）assisted cascade decoding algorithm is proposed to reduce the complexity of the BP-OSD algorithm.The algorithm achieves the purpose of adjusting the decoding complexity and decoding performance by dividing the CRC into two segments,which are used as code word check and error correction respectively.The simulation results show that the balance between complexity and performance can be maintained by adjusting the CRC length.3.An improved algorithm that cascades BPL with OSD is proposed.When the BPL decoder reaches the maximum number of iterations without passing the checksum,the OSD decoder starts.Two cascade structures,series and parallel,are also proposed.Simulation results show that both structures can significantly improve the performance.4.Controllable artificial noise is proposed to be introduced in the BPL-OSD algorithm to reduce oscillations.Two types of noise are introduced for this purpose:fixed noise and variable noise.The standard deviation of the fixed noise is fixed under each path,while the variable noise increases with the number of paths in which it is located.Simulation results show that the algorithm with fixed noise assistance has a performance gain of 0.5 d B compared to the algorithm without noise assistance,and the performance of the algorithm with variable noise has a performance gain of 0.01 d B compared to fixed noise.Both outperform the CRC Aided Successive Cancellation list（CASCL）algorithm at a list size of 8for short code lengths.