Font Size: a A A

Implementation Of Deep Learning Technology For Massive MIMO Signal Detection

Posted on:2022-06-10Degree:MasterType:Thesis
Country:ChinaCandidate:P JiaFull Text:PDF
GTID:2518306521994779Subject:Electronics and Communications Engineering
Abstract/Summary:PDF Full Text Request
Massive MIMO is one of the key technologies to improve the efficiency 5g spectrum.However,the increase in the number of antennas in Massive MIMO system causes the channel matrix dimension to become larger,and traditional signal detection methods face great challenges.The deep network structure of deep learning makes it have strong data learning ability and can realize the approximation of complex functions.In this paper,the application of deep learning technology in Massive MIMO signal detection is studied.The particular research work and innovation are as follows:Firstly,the structure of Det Net detection network is ameliorated to improve the detection performance.Firstly,the detection principle of Det Net is deeply analyzed,and the structure of each detection unit in the network is simplified by removing redundant inputs;Secondly,in order to further improve the performance of the network,the idea of random forest is used to extend the detection unit of each layer of the original network from one to two,and a twin network is constructed,and the different initial value vectors are assigned to the two detection units.The simulation results show that the performance of the improved network is better than that of the original Det Net.Secondly,a Massive MIMO signal detection scheme based on ZF algorithm and fully connected neural network is proposed.Firstly,a deep neural network is constructed by using the full connection layer.Then,based on the deep neural network,combined with the traditional ZF detection algorithm,a new signal detection scheme is established.The BER performance of ZF algorithm,fully connected neural network and ZF-DNN signal detection scheme proposed in this paper are analyzed by simulation experiments.The data results show that ZFDNN has better BER performance,and the BER performance is still stable even in the relatively bad channel environment.The research in this paper shows that deep learning has great advantages in Massive MIMO signal detection,which provides a new way for the future research of wireless communication system.
Keywords/Search Tags:Massive MIMO, Signal Detection, Deep Learning, Duplication Network, Zero Forcing algorithm
PDF Full Text Request
Related items