| Orthogonal Frequency Division Multiplexing and Multiple Input Multiple Output technology have been one of the key technologies in the fourth and fifth generation of mobile communication system,which can improve the channel capacity and have good anti-interference ability.When the number of users and antennas in the communication system increases,the computational complexity of channel estimation and signal detection will increase significantly.In order to achieve high detection accuracy under the condition of low complexity,channel estimation and signal detection based on deep learning method are studied in this paper.Aiming at the problem of channel estimation in MIMO-OFDM communication system,a channel estimation method based on image denoising network is proposed.Firstly,the least square channel estimation algorithm is used to obtain the initial channel estimation,in which the error between the estimated result and the actual channel exists.The image denoising neural network is used to remove the estimation error,and the asymmetric loss function is added to the loss function to improve the performance of the channel estimation.For the pilot power allocation problem in multi-user MIMO-OFDM communication system,a joint optimization algorithm of pilot allocation and channel estimation is proposed.Under the condition of given total pilot power constraint,the full connect neural network and channel estimation network are cascaded to realize pilot power allocation and channel estimation.For the signal detection problem,the zero forcing detection algorithm is used to obtain the initial detection signal,and then the signal is predicted by the fully connected neural network to improve the signal detection performance of the system.In order to verify the effectiveness of the algorithm,the proposed algorithm is compared with the traditional detection algorithm by simulation.The results show that,compared with the traditional algorithm,the detection algorithm based on deep learning has higher detection accuracy under the condition of low computational complexity. |
