Research On Detection Of Dynamic Of Interference User In Downlink Non-Orthogonal Multiple Access Systems

In recent years,people spend more time on the mobile Internet,and the In-ternet of Things(IoT)has also flourished.And many intelligent IoT terminals,which need wireless access,have been researched and developed.Thus,peo-ple have put forward higher requirements for wireless communication speed.Recently,the fifth generation of mobile wireless communication technologies mainly face the following challenges,including high throughput,low latency,high reliability,and large connections.As an important candidate in 5G,non-orthogonal multiple access technology has received extensive attention from academia and industry.Non-orthogonal multiple access technology has mul-tiple implementations,and non-orthogonal in power domain is one of them.Users with extremely different channel conditions are more likely to share the same time-frequency resource block,and each user is assigned a power ratio to ensure that the total power is constant.The receiver extracts the target signal by using a serial interference cancellation technique.However,in order to cor-rectly extract the target signal,the receiver needs to know modulation order of the interference user.In this paper,the blind detection method of modulation order of interference user in non-orthogonal multiple access systems is studied.Firstly,this paper studies the existing maximum likelihood blind detec-tion algorithm and the maximum log likelihood blind detection algorithm,and points out the advantages and disadvantages of these two algorithms.In or-der to reduce the computational complexity of the maximum likelihood algo-rithm and improve the blind detection performance of the maximum log like-lihood algorithm,this paper considers the statistical characteristics of the re-ceived signal,optimizes the maximum log likelihood algorithm,and proposes the K-maximum log likelihood algorithm.This paper analyzes and compares the detection accuracy and throughput performance of different algorithms.The simulation results show that the detection accuracy and throughput performance of the K-maximum log likelihood algorithm are better than the maximum log likelihood algorithm.And its performance is close to the ideal case in some scenarios.In addition,the complexity of the maximum likelihood algorithm,the maximum log likelihood algorithm and the K-maximum log likelihood al-gorithm are compared.It is found that the computational complexity of the K-maximum log likelihood algorithm is much lower than the maximum likeli-hood algorithm and is very close to the maximum log likelihood algorithm.Secondly,this paper firstly proposes the characteristics of interference user modulation order based on Anderson-Darling test,introduces machine learning algorithm into the field of blind detection,and proposes MLAD algorithm.The MLAD algorithm is mainly composed of five steps,including clustering,fea-ture extraction,classification model training,model parameter selection and decision.The feature extraction phase uses the Anderson-Darling test,and the feature mapping process uses a polynomial mapping.In this paper,the perfor-mance of MLAD algorithm is simulated and compared with the performance of maximum log likelihood algorithm and K-maximum log likelihood algorithm.From simulation results,we can conclude that the performance of the MLAD al-gorithm is the best among the three algorithms.And its performance is close to the ideal situation in most scenarios,even the same as the ideal case.In addition,this paper also studies the performance of MLAD algorithms with different ma-chine learning algorithms.The simulation results show that the performance of MLAD algorithms are almost the same when using logistic regression,support vector machine and artificial neural network.Therefore,in practical applica-tions,MLAD can adopt a machine learning algorithm with low complexity and easy implementation.