Blind Recognition of LDPC Codes by Average LLR in Multi-Channels Based on CNN

Coding and decoding play an essential role in communication systems and are the primary conditions to ensure the standard transmission of signals, which can increase stability and reliability after coding. Low Density Parity Check Code (LDPC) coding is a common form of coding, and its parallel iterative decoding algorithm based on a sparse matrix is less complex and more flexible than the Turbo algorithm. However, with the increasing number of protocols and channels, the traditional methods of LDPC codes recognition are not adaptable enough, a more flexible and applicable method is essential. In this paper, we propose a deep learning-based LDPC coding recognition method, which is more adaptable, has higher estimation accuracy than traditional algorithms and has a more straightforward data processing structure compared to 2-dimensional convolutional neural networks and fully connected networks. After training the data of the Additive White Gaussian Noise (AWGN) channel, Rayleigh channel and Rician channel, the network's performance is better in the test, and the estimation accuracy is better than the traditional algorithm.