RS code belief propagation decoding method based on deep learning

Abstract

The invention relates to an RS code belief propagation decoding method based on deep learning, and the method comprises the following steps: building a non-full connection neural network through employing a deep learning method according to a Tanner graph corresponding to an RS code parity check matrix; converting an operation process of a check node and a variable node in the Tanner graph into anoperation process of neurons in the neural network; initializing a parity check matrix parameter to be 0 or 1 to serve as a weight value for training, and using a deep learning optimized parameter value in a corresponding Tanner graph, namely a variable node layer in a corresponding neural network, namely a variable node layer for neural network training; inputting a bit-level log likelihood ratio (LLR) received from the noise channel into the neural network as reliability information; performing summation operation on the LLR value; and performing symbol-level displacement of random length on the codeword after each iterative computation by using an SSID algorithm.

Description

[0001] Technical field [0002] The invention belongs to the field of error control coding in channel coding, and relates to a Reed-Solomon code (RS code) belief propagation soft decision decoding algorithm using deep learning technology. Background technique [0003] In recent years, with the development of the information society and the continuous improvement of communication technology, people's requirements for the reliability of data transmission are increasing day by day. How to ensure the reliable transmission of data has become one of the issues that must be paid attention to in the design of communication systems. Since Shannon proposed the theory of channel coding in 1948, the application of error control coding has become a research hotspot in modern communication systems and storage systems. When the sent information is sent from the source to the receiver, it will cause random errors in the information due to the unsatisfactory transmission channel during the tra...

AI Technical Summary

Problems solved by technology

This algorithm uses the bit-level log likelihood ratio (Log Likelihood Ratio, LLR) as the input data of the decoder, and uses a coefficient to improve the performance of the code after summing the outer LLR values ​​of this iteration at the output end. It is called the damping coefficient (Damping Coefficient), but the setting of this value is set using the empirical value obtained from the simulation, and there is no accurate mathematical derivation

Images