Research On Channel Estimation And Signal Detection Of MIMO System Based On Deep Learning

As the rapid increase of business types and data volume of wireless communication system,the demand for system capacity limit and processing speed is growing.Multiple input multiple output(MIMO)technology can achieve higher channel capacity and rate without increasing system bandwidth and transmit power.However,channel estimation and signal detection of MIMO wireless communication system are more difficult than single antenna system.Therefore,the research on this aspect is still a hot spot in the field of wireless communication.In addition,deep learning is widely used in the network layer and application layer of communication system.With the continuous updating of technology,it is advancing to the physical layer of communication system.Therefore,based on deep learning,channel estimation and signal detection of MIMO system are studied in this thesis.Accurate channel estimation can have an important impact on the performance of wireless communication system.Aiming at the problem that traditional channel estimation algorithms for MIMO system need to know channel statistics,a channel estimation scheme based on deep learning is proposed.Firstly,the channel information reconstruction network is constructed based on the convolutional neural network(CNN),and the channel information is preliminarily reconstructed.Then the channel estimation network is constructed based on the residual deep neural network(Res-DNN)in cascade,and the final estimation result is obtained.In the proposed scheme,multiple loss functions are used to optimize the network,and the channel information is reconstructed before channel estimation to estimate the channel more accurately.Simulation results show that the performance of the proposed scheme is better than that of the linear minimum mean square error(LMMSE)estimation algorithm with the increase of SNR,and the channel information is not needed to be known.Designing a suitable signal detection algorithm at the receiver can give full play to the potential of MIMO.Aiming at the tradeoff between performance and complexity in traditional signal detection of MIMO system,an end-to-end signal detection scheme based on res DNN is proposed.The encoder and decoder based on res DNN replace the transmitter and receiver of wireless communication system respectively,and train through the end-to-end way.Firstly,the encoder extracts the features of the input data,then establishes the communication model and sends it to the zero forcing detector for preliminary detection.Finally,the decoder reconstructs the detection signal.The end-toend training mode avoids the influence between modules and improves the system efficiency to a certain extent.Simulation results show that the detection performance of the proposed scheme is better than that of the same type of algorithm,and it is significantly better than the minimum mean square error(MMSE)detection algorithm at the expense of a certain time complexity.