MULTI-SACLE 1D-CNN DEMODULATOR FOR LOW OVERSAMPLING DIGITAL MODULATION SIGNAL
Aiming at the problem of high oversampling requirements when applying deep learning methods to demodulate of digital modulation signals,this paper designs a multi-scale one-dimensional convolutional neural network digital demodulator with low oversampling.It could demodulate the four digital modulation signals of BPSK,4-QAM,8-QAM,and 16-QAM under the same oversampling conditions as the traditional demodulator,and could ensure the same error performance of the traditional demodulation method.Simulation results show that under Gaussian and Rayleigh fading channels,the provided digital modulation signal demodulator can not only ensure the performance of demodulation error codes,but also reduce the requirement of sampling multiple,and also reduce the complexity of neural network structure.
Low sampling multipleDemodulationMulti-sacle 1D-CNNBPSK and M-QAM
国家科技期刊平台
NSTL
万方数据
