Polarization code decoding method based on deep neural network

Abstract

The invention provides a polarization code decoding method based on a deep neural network. The method comprises the steps of: firstly, collecting and arranging sample data; secondly, setting parameters for modeling, and adopting a backpropagation algorithm to train the network; thirdly, inputting a likelihood ratio corresponding to a Rate-R node into the trained deep neural network model to obtain0 or 1; and finally, executing a simplified successive elimination decoding algorithm according to the 0, 1 state. By combining the deep neural network technology with the polarization code decodingtechnology, the polarization code decoding method reduces the traversal operation of the Rate-R node, accelerates the decoding speed, and reduces the decoding delay.

Description

technical field [0001] The invention belongs to the technical field of communication, and in particular relates to a fast decoding with a deep neural network to assist a simplified continuous elimination decoding algorithm. Background technique [0002] Polar code is a new type of channel coding proposed by E.Arikan in 2008. Polar codes are the first constructive coding schemes that can be proved to achieve channel capacity through rigorous mathematical methods, and have clear and simple coding and decoding algorithms. Through the continuous efforts of channel coding scholars, the error correction performance that the current polar codes can achieve exceeds that of the currently widely used Turbo codes and LDPC codes. [0003] The basis of polar codes is the channel polarization. When the number of channels (timeslots) participating in the channel polarization is large enough, the channel capacity of the obtained polarized channel will have a polarization phenomenon, that ...

AI Technical Summary

Problems solved by technology

Although SC decoding has low complexity and simple decoding structure, it can only decode bit by bit, which brings high delay

Images