Polar code SSCL algorithm decoder based on deep learning

Abstract

The invention provides a low-time-delay polar code decoder combined with deep learning, and particularly, compared with a simplified Successive cancellation list (SSCL) decoder, the low-time-delay polar code decoder has the advantage that the number of calculation times is smaller. The decoder comprises five modules, namely a normal SCL calculation module, a Rate-0 calculation module, a Rep calculation module, a Rate-1 calculation module, and a DNN calculation module of a common node. The SSCL decoder combined with the deep neural network retains the original low decoding delay characteristicof Rate-0, Rate-1 and Rep nodes, a deep neural network is used for decoding common nodes to reduce the decoding delay, and finally the purpose of reducing the overall decoding delay is achieved. Experimental calculation shows that when the code length is 64 and the code rate is 1 / 2, the decoding delay is reduced by about 27% compared with the decoding delay of the SSCL.

Description

technical field [0001] The invention belongs to the technical field of decoding of communication channel coding, and relates to a polar code decoder combined with deep learning, in particular, a decoder with fewer calculation times compared with the traditional SSCL algorithm, reaching A Polar Code Decoder Using Deep Learning to Reduce Decoding Latency Effects. Background technique [0002] Since the establishment of channel coding theory, channel coding technology has experienced decades of development and innovation. As a coding technology proposed in the past ten years, polar codes have a specific codec structure like algebraic codes, and also use channel polarization to establish a theoretical basis for codecs. Polar codes obtain several split channels through two operations of channel combination and channel splitting, and their capacity shows a trend of polarization, that is, as the code length increases, either tends to a completely noisy channel, or tends to a compl...

AI Technical Summary

Problems solved by technology

[0005] In order to solve the problem that SSCL has a higher delay for nodes other than special nodes, the present invention provides a decoder combined with the SSCL decoding algorithm of deep learning neural network, which mainly includes 5 calculation modules, normal SCL calculation module, Rate -0 calculation module, Rep calculation module, Rate-1 calculation module and DNN calculation module of ordinary nodes

Images