MSK Demodulator and Impulsive Noise Depression Based on Convolutional Neural Network with Gated Layers

In this paper, a novel demodulation scheme is proposed to demodulate minimum shift keying (MSK) signals and depress the impulsive noise. The combination of convolutional neural network (CNN) and GatedNet is able to achieve both demodulation and impulsive noise depression, which distinguishes the proposed scheme from others. Furthermore, considering the features of demodulation, the use and structure of GatedNet is redesigned in this demodulation scheme. The simulation results demonstrate that this scheme can improve demodulation performance about 2dB under impulsive noise, compared with the demodulation based on coherent sequence detection and an impulsive based branch metric, whose performance can closely approach the performance of maximum likelihood algorithm.