Research On Detection Method For MIMO-OFDM Systems Based On Deep Learning

Orthogonal Frequency Division Multiplexing(OFDM)technology has been widely studied and applied because of its high spectral efficiency and effective resistance to frequency selective fading.However,in practical applications,the length of the channel impulse response may exceed that of the cyclic prefix(CP),which is called CP insufficiency.Insufficient CP will lead to the existence of inter-carrier interference(ICI)and inter-symbol interference(ISI)in the received signals,so that the bit error rate of single tap detection algorithm in the high-order modulated OFDM systems increases sharply,and the receiver cannot work.Therefore,it is necessary to study the OFDM systems with insufficient CP and propose high performance detection algorithms with low complexity suitable for high-order modulated OFDM systems.Firstly,the application of the block multi-subcarrier joint detection based serial interference cancellation algorithm in the detection of single input single output(SISO)OFDM systems with insufficient CP is studied.To solve the high bit error rate of classic serial interference cancel-lation algorithm,a block maximum likelihood based serial interference cancellation algorithm and a block Det Net based serial interference cancellation algorithm are proposed.The joint de tection of subcarriers improves the accuracy of the decision in every iteration,reduces the error propagation in the interference cancellation process,and finally realizes the improvement of the detection performance.The simulation results show that the detection performance of the proposed block multi-subcarrier joint detection based serial interference cancellation algorithm can approach or even exceed the bit error rate performance of OFDM systems with sufficient CP under some simulation scenarios.And the block Det Net based serial interference cancel-lation algorithm shows good robustness for different channel power delay profile and different number of subcarriers in OFDM systems.Then,a convolutional neural network(CNN)based parallel inter ference cancellation al-gorithm is proposed for SISO-OFDM systems with insufficient CP.Compared with the block multi-subcarrier joint detection based serial interference cancellation algorithm,the CNN based parallel inter ference cancellation algorithm can realize the joint detection of more subcarriers.Moreover,a Dense Net based parallel inter ference cancellation algorithm is obtained by using the optimized network architecture Dense Net to replace the CNN part in the CNN based parallel inter ference cancellation algorithm.The simulation results show that when the length of CP is0,compared with the block Det Net based serial interference cancellation algorithm the CNN based parallel inter ference cancellation algorithm and the Dense Net based parallel inter ference cancellation algorithm can further improve the detection performance of OFDM systems with insufficient CP.Moreover,the simulation of frame error rate in the coding environment shows that with the help of fully connected network,the block Det Net based serial interference cancel-lation algorithm and the Dense Net based parallel inter ference cancellation algorithm can output soft information which can be used by the channel decoder.Finally,the CP insufficiency problem is extended to the multiple input multiple output(MIMO)OFDM systems.The block Det Net based serial interference cancellation algorithm is improved to make it suitable for MIMO-OFDM systems with insufficient CP.The simulation results given in this paper show that in a CP-free MIMO-OFDM system,the block Det Net based serial interference cancellation algorithm can achieve a detection performance that completely surpasses the bit error rate performance of MIMO-OFDM system with sufficient CP.Moreover,the block Det Net based serial interference cancellation algorithm can adapt to the change of MIMO channel correlation without the need to train the network repeatedly.