Improvement of Iterative Decoding with LLR Modulator by Neural Network Using Magnetic Transition Information in SMR System

We have previously focused on a log-likelihood ratio (LLR) computed as the decoding reliability by $a$ posteriori probability (APP) decoder on the low-density parity-check (LDPC) coding and iterative decoding system for the shingled magnetic recording (SMR). Then, we clarified that the LLR modulator applying a neural network realizes an effective iterative decoding. In this article, we propose a new neural network LLR modulator considering magnetic transition information obtained from LLRs to reduce the influence of intersymbol interferences (ISIs).