Joint Channel Estimation And Active User Detection Based On Compressed Sensing In NOMA System

In the 5G period,due to the rapid development of mobile communication,spectrum resources are even more scarce.In order to meet the rapidly increasing requirements of mobile services,people have begun to seek new mobile communication technologies that can not only meet the experience requirements,but also improve the spectrum efficiency.In this context,non orthogonal multiple access(NOMA)technology has obtained great attention.NOMA technology is mainly divided into two categories: power domain multiplexing and code domain multiplexing.Power domain multiplexing NOMA has completely changed the previous way of monopolizing resources by a single user in the power domain,and clearly put forward the idea that power can also be shared by many users;For code domain multiplexing,many research institutions have proposed different implementation schemes whose ultimate goal is to improve the spectrum utilization,just like power domain NOMA.In addition,NOMA technology can completely distinguish each user through interference elimination technology at the receiving end.The key business application of 5G is mobile network and Internet of things,and the general scenario of Internet of things connection includes a large number of machine type network connection devices.In typical applications,only a small number of(unknown)devices are active at a given time,presenting a certain sparsity.Since 5G system needs to meet the characteristics of low delay,the traditional authorization based random access scheme is no longer applicable.A promising solution is the unauthorized random access scheme based on advanced compressed sensing(CS)technology.Compressed sensing is an attractive technology to collect,process and recover sparse information.This paper mainly carries out channel estimation(CE)and active user detection(AUD)in uplink scheduling free NOMA system based on CS model.The main research work and contents are as follows:1)According to CS principle,an improved approximate message passing(AMP)algorithm,threshold adaptive constrained reweighting approximate message passing(TA-CR-AMP),is proposed to deal with the problem of uplink CE and AUD in the scenario of multi-user information transmission in NOMA system.Simulation results show that,as compared with other algorithms,the proposed algorithm can obtain lower mean square error(MSE),higher successful activity detection rate(SADR)and faster convergence.2)Aiming at the problem of multiple measurement vector(MMV),according to the distributed compressed sensing mechanism and the structural sparse feature of the reconstructed vector,A novel compressed sensing reconstruction algorithm,multiple measurement vector threshold adaptive constrained reweighted approximate message passing(MMV-TA-CR-AMP),is proposed to solve the joint CE and AUD problem of multi-user transmission scenarios in SIMO-NOMA system.Simulation results show that this algorithm can obtain better MSE,SADR and convergence performance than other algorithms.3)This paper optimizes the recovery matrix in compressed sensing by minimizing the crosscorrelation value of the columns in recovery matrix,and explores the impact of recovery matrix optimization on the two new proposed algorithms in this paper,so as to deepen the understanding of the improvement principle of the new algorithm.Simulation results show that the optimization of recovery matrix will degrade the performance of TA-CR-AMP and MMV-TA-CR-AMP algorithms with threshold adaptation.The main reason is that the optimization of recovery matrix will make the threshold rise too fast in the threshold adaptation operation,which deteriorates the performance of the algorithm.For the two new algorithms CR-AMP and MMV-CR-AMP removing threshold adaptation,the optimization of recovery matrix not only slightly improves the performance,but also significantly reduces the running time of the algorithm,and the performance and operational complexity are better than those of TA-CR-AMP and MMV-TA-CR-AMP algorithms without optimized recovery matrix.