A Deep Learning Based Method for Blind Recognition of LDPC Codes

Abstract

Deep learning is an emerging research direction in machine learning, which has a promising application in the field of communications. In this paper, we focus on adaptive coding systems based on LDPC codes and study the problem of blind recognition with a pre-defined LDPC encoder candidate set. We propose a deep learning (DL)-based method for blind recognition using the log-likelihood ratios (LLR) of the syndrome a posteriori probabilities (SPP). The proposed method requires only a simple Multi-Layer Perceptron (MLP) and can therefore be easily used for systems with high real-time requirements as well as can be easily adapted to different codes and channel parameters. Simulation results show that the proposed approach allows for a comparable recognition performance to existing methods.