Research On Low-complexity Detection For Sparse Code Multiple Access System

This thesis discusses the low-complexity multi-user detection algorithm of Sparse Code Multiple Access system(SCMA).In SCMA system,according to the allocated user-specific codebook,multi-dimension codewords are multiplexed into the same resource blocks as an implementation of non-orthogonal multiple access.The non-orthogonality of SCMA acheives higher spectrum efficiency,but the receiver needs to support the multi-user detection algorithm with higher complexity to assure the bit error rate(BER)performance.The sparsity of SCMA makes Message Passing Algorithm(MPA)available in receiver.MPA has approximately optimal BER performance,while it is too complex to implement since its large number of exponential operations.So this thesis focus on the research of low-complexity detection algorithm for SCMA receiver,which mainly consists of the following four parts.Firstly,the SCMA system model and fundamental are given.The system's specific constitutes and principles are separately studied in the transmitter and receiver.On the base of that,the downlink system model is described.Besides that,the mechanism of original MPA detection and the reason for its high complexity are described in detail.To decrease the exponential operations in MPA,the logarithmic domain MPA detection is also introduced,which is denoted as Log-MPA.Secondly,the incompletely iterated MPA detection is proposed.By analyzing the high complexity of original MPA,it is found that by implementing incompletely iterated MPA to some users,the total detection complexity can be largely reduced.Specifically,two algorithms are proposed based on this principle,which are Multi-stage Message Passing Algorithm and Threshold-based Message Passing Algorithm.It is further demonstrated by simulation that the proposed algorithms can achieve a much lower complexity with small performance loss and good convergence compared to the original MPA.Thirdly,sequential MPA detection is discussed and it is further combined with the incompletely iterated MPA detection.In sequential MPA,the inter-iteration message exchange between the user nodes is introduced,which leads to faster convergence by less iterations and further reduction in detection complexity.Moreover,sequential MPA can be combined with incompletely iterated MPA to further decrease detection complexity and compensate the BER performance loss,which is proved by simulation.Finally,Successive Interference Cancellation(SIC)is introduced to SCMA system to pursue the decrease of complexity in the order of magnitude.By remodeling system and designing specific codebook,this thesis gives the SIC detection applied for SCMA system.Since there are no exponential operations,but only some subtracts in SIC detection,its complexity can be reduced in the order of magnitude compared to MPA.However,there is some loss in the BER performance compared with MPA.To improve the performance of SIC,this thesis presents SIC-MPA detection by combining SIC with MPA.It is demonstrated by simulation that SIC-MPA has a better BER performance than SIC with much lower complexity than original MPA.