Research On Deep Machine Learning Detection Method For MIMO/OFDM System

In recent years,the artificial intelligence method based on deep machine learning has been widely used in the field of mobile communication because it can make up for the discrepancy between the algorithm based on the statistical model of system elements and the actual scene.This paper attempts to apply the deep machine learning method to the signal detection of MIMO and OFDM systems,and designs the corresponding deep learning detection method.The main work is as follows:1)Aiming at the problem of MIMO signal detection under the condition that CSI is known,an improved Det Net deep learning detection method is designed.For the simplified Det Net detection network,some redundant input vectors are deleted,and the parameters of the detection network are trained.The simulation results show that the BER performance of the improved Det Net detector is better than that of the traditional Det Net detector,and the BER performance of Det Net is also better than that of the traditional ZF detector.2)Aiming at the problem of MIMO signal detection under the condition of unknown CSI,the DNN architecture based on deep learning is adopted,and the deep learning DNN detector is obtained by training the network parameters.Simulation results show that the BER performance of the proposed DNN detector is very close to that of the ML detection algorithm under perfect CSI conditions without channel estimation.3)Aiming at the problem of OFDM signal detection under the condition that CSI is known,an improved Det Net deep learning detection method is designed.A simplified Det Net detection network is proposed,some redundant input vectors are deleted,and the parameters of the detection network are trained.The simulation results show that the performance of the improved Det Net detector is better than that of the traditional ZF detection even in the case of large Doppler frequency offset.4)Aiming at the problem of OFDM signal detection under the condition of unknown CSI,the deep learning DNN network is adopted,and the deep learning DNN detector is obtained by training the network parameters.Aiming at the existing OFDM transmission system with pilot,the DNN detector takes the pilot and data as the input of the network at the same time,which improves the detection performance of the network.Simulation results show that the BER performance of the designed deep learning DNN detector is similar to that of the traditional MMSE detection algorithm.However,in the case of small number of pilots,removal of cyclic prefix and nonlinear clipping noise,the BER performance of DNN algorithm is better than that of traditional MMSE detection methods.