Research On Multi-User Blind Detection And Its Parallel Processing For Sparce Code Multiple Access

To meet the challenging demand of the future 5G communications,Sparse code multiple access(SCMA)is proposed to be used in the air interface of 5G networks.As a non-orthogonal multiple access,SCMA can make a great breakthrough of conventional orthogonal multiple access' s overhead and support 5G requirement of massive connectivity.In uplink scenario,SCMA can be combined with contention-based grant-free transmission mechanism to support massive and concurrent random access.To do so,the receiver of uplink grant-free SCMA is supposed to detect the active users and decode the transmitted data without prior information of pilot sequences and codebooks.Therefore,this paper focus on the blind detection algorithms used in the uplink grant-free SCMA receiver.First,this paper shows the limitations of existing mobile communication system,and a solution that combines the contention-based random access and non-orthogonal multiple access is proposed.Then the system model of uplink grant-free SCMA and the structure of its receiver are established.Next,an algorithm for active user detection based on the framework of sparse Bayesian learning(SBL)is proposed and studied.The algorithm can detect the active users and estimate their corresponding channel without prior knowledge of user sparsity.What's more,computation complexity of the proposed algorithm is evaluated,and numerical simulation results are provided to substantiate and further analyze the performance of this algorithm.Finally,the message passing algorithm(MPA)used for multi-user detection and data decoding is introduced,and an innovative parallel-processing strategy of MPA based on general-purpose computing on graphics processing units(GPGPU)is proposed according to the parallel manner of MPA's message update and the requirement of soft baseband.Functional simulation of the parallel-processing strategy is implemented on a graphics processing units(GPU),and the results of decoding efficiency demonstrate that the strategy can considerably accelerate the decoding process and reduce the latency.Meanwhile,the launch details and the performance analysis of the key kernel functions are presented in expectation of making further improvement on current scheme.