Research On Random Access System Based On Dictionary Learning In MMTC Scenario

With the popularization of the fifth generation(5G)mobile communication and the vigorous development of the Internet of Things technology,people’s demand for communication quality is increasing day by day.The deployment of massive devices and the rapid change of Internet traffic make the information data show an explosive growth trend,which requires wireless communication to have the ability of large connection and wide coverage to meet the device access problem in this scenario.In order to solve the problem of massive access conflicts and improve the accuracy of channel estimation and pilot detection,this paper uses a dictionary learning algorithm to sparsely represent the channel,and uses this sparse representation design algorithm to perform joint channel estimation and pilot detection.Research work includes:(1)Aiming at the characteristics of channel sparse structure in massive Machine-Type Communication(mMTC)scenarios,a method for sparse representation of channel blocks based on dictionary learning is proposed,and the effects of sparse block length,redundancy factor and number of samples on channel sparse representation are discussed.Firstly,the wireless random access scenario is described by the Geometry-based Stochastic Channel Model(GSCM),and the dictionary is improved by considering the block sparse structure in massive multiple-input multiple-output(massive MIMO)systems.The K-Singular Value Decomposition(K-SVD)algorithm in learning makes a sparse representation with a block structure for the original channel.The simulation results show that the sparse representation accuracy of the channel is positively correlated with the length of the sparse block and the size of the redundancy factor;the greater the number of samples,the lower the channel sparse representation accuracy.(2)Aiming at the problem of limited pilot resources in massive access,a cluster analysis algorithm of the same size is proposed,and a pilot multiplexing method is proposed based on the algorithm.The users in the cell are divided into several sectors by clustering,and the number of users in each sector is limited to be the same,so as to ensure that the user channel difference between sectors is relatively large.Reduces the pilot requirements for the entire cell to the pilot requirements for one sector.Comparing the traditional clustering with the same-sized clustering and the reference model,it can be found that the same-sized clustering method can more effectively divide the users in the cell into equal number of sets,and users in different sets can reduce the number of cells by sharing pilot frequencies.The number of pilots needs to be saved,thereby saving pilot resources.(3)Aiming at the block sparse channel characteristics caused by the angle expansion of the mMTC scene,a block sparse compressed sensing joint channel estimation and pilot detection algorithm based on dictionary learning is proposed.The algorithm uses the dictionary obtained by dictionary learning and training,and considers the characteristics of angle spread in wireless channel scenarios,and proposes a block-sparse-based compressed sensing algorithm,thereby performing joint channel estimation and pilot detection.The performance of channel estimation and pilot detection generally improves as the number of active devices decreases,the pilot length,and the signal-to-noise ratio increase.Research shows that the performance of the proposed algorithm in channel estimation and pilot detection is significantly better than other algorithms.This paper conducts an in-depth study on the random access scenario of mMTC,using the block sparse structure formed by angle expansion,using the compressed sensing method with a specific sparse structure,and the highly sparse domain constructed by dictionary learning effectively improves the channel estimation and pilot detection.At the same time,the same size clustering is used to realize user partitioning,so as to achieve the purpose of pilot frequency reuse,thereby saving pilot frequency resources.