Research On Signal Detection And Optimization Scheme Of MIMO Systems Based On Deep Learning

At present,as one of the main research directions in the field of wireless communication,intelligent communication introduces actively artificial intelligence technology into all levels of wireless communication system,which has become an effective way to explore the field of intelligent communication.Because of the increase of the number of antennas configured in the receiver and transmitter of Multiple input multiple output(MIMO)systems,the process of signal processing in the communication systems receiver will become more complex.In order to achieve a better compromise between the performance and complexity of MIMO systems signal detection,this thesis studies deep learning algorithm to jointly solve the problem of MIMO systems signal detection and channel decoding.At the same time,the deep learning method is proposed to jointly optimize the functional modules of the physical layer of traditional MIMO systems.The main research contents of this thesis are summarized as follows:1.A deep learning method is proposed to solve the problems of signal detection and channel decoding in MIMO systems.First of all,the MIMO systems model based on neural network are constructed.Deep neural network(DNN)and autoencoder(AE)neural network are respectively introduced into the traditional MIMO systems receiver in order to obtain the transmission information bits or codewords and channel state information.Then,the DNN and AE network model are trained by an end-to-end way,so that the neural networks can learn the mapping relation the transmission information bits or codewords of the receive and transmitter in order to realize jointly the signal detection and channel decoding of the MIMO systems with low complexity.Thirdly,the performance of the trained network model is tested.Finally,the simulation results show that the proposed method can improve the performance of detection and decoding.2.A joint optimization scheme of MIMO systems based on deep learning is proposed.Aiming to joint optimization problem of physical functional modules in traditional MIMO systems,the thesis studies the deep learning algorithm to solve the whole optimization problem of MIMO systems.Firstly,the autoencoder neural network is proposed to construct a multi-user communication systems model.The transmitter and receiver of traditional MIMO systems are respectively regarded as the encoder anddecoder of the autoencoder neural network.Then,the method of cross entropy loss weighted sum function is used to train the system model to obtain a system model with optimal transmission performance.In a certain range of signal-to-noise ratio(SNR),the trained model is tested to acquire the bit error rate of each user and the average bit error rate of the whole multi-user.Finally,the simulation results show that the MIMO communication systems based on the autoencoder have better system performance than the traditional MIMO systems.