GatedNet: Neural Network Decoding for Decoding Over Impulsive Noise Channels

This letter proposes a novel neural network (NN) called GatedNet for decoding over-impulsive noise channels. To reduce the impact of impulsive noise, the neurons in the GatedNet are redesigned with control gates to achieve better performance and robustness. Furthermore, a partially connected layer (PCL) in the neural network design is proposed to reduce the computational complexity. Simulation results show that the proposed GatedNet decoder outperforms the belief propagation (BP) decoder for different impulsive noise channels. Furthermore, it achieves near-maximum likelihood (ML) performance with much shorter decoding time.