Non-Iterative MSNN Decoder for Image Transmission over AWGN Channel using LDPC

The iterative nature of existing Low-Density Parity-Check (LDPC) decoding algorithms poses a significant obstacle to achieving fast reconstruction of received images with a low error rate. These decoders rely on probabilistic values, which necessitates an iterative approach. In this study, a novel approach to decoding LDPC codes that involves using a multilayer shallow neural network (MSNN) is proposed. By leveraging machine learning algorithms, the MSNN decoder avoids the iterative nature of existing decoders. LDPC construction Mackay Neel (MN) LDPC construction and Quasi Cyclic (QC) has been adopted to transmit the image over Additive White Gaussian Noise (AWGN). The performance of MSNN decoder against existing algorithms such as Bit-Flipping (BF), Belief Propagation (BP), and Sum-Product Algorithm (SPA) has compared. The experiments demonstrate that the MSNN decoder outperforms existing decoders by approximately 10 2 at 3 dB Signal-to-Noise Ratio (SNR). Evaluated the reconstructed image using Peak Signal-to-Noise Ratio (PSNR) and Entropy for both SPA and MSNN decoders. The results show that the MSNN decoder performs exceptionally well, even under qualitative analysis of the reconstructed image. Overall, the findings suggest that the proposed MSNN decoder offers a promising solution for fast and accurate reconstruction of received images with low error rates.