Research On Signal Processing Algorithm Of Non-Orthogonal Multiple Access System

Non-orthogonal multiple access(NOMA)is a key technology for the development of the fifth generation mobile communication.The spectrum efficiency is effectively improved by multi-user multiplexing in the power domain.Different power allocation schemes directly affect the throughput of the system,and because the non-orthogonal multiple access technology is utilized at the transmitting end of the NOMA system,the difficulty of signal interference cancellation at the receiving end is increased.Therefore,power allocation and interference cancellation are two key technologies of non-orthogonal multiple access communication systems.This paper mainly studies the power allocation algorithm of the transmitting end and the interference cancellation algorithm of the receiving end in the non-orthogonal multiple access system.Firstly,aiming at the local optimal problem of NOMA downlink power allocation algorithm at the transmitting end,an optimal power allocation scheme using conjugate gradient method is proposed,which is used to solve the user's maximum weighted sum rate.The existing theory proves that the conjugate gradient method can converge to the global optimal solution.The simulation results show that its performance is better than fixed power allocation(FPA)algorithm's and fractional transmit power allocation(FTPA)algorithm's,and NOMA system with this method is better than orthogonal multiple access(OMA)system.Secondly,the interference cancellation algorithm that is desired at the receiving end not only has low algorithm complexity,but also has a low bit error rate.The parallel interference cancellation(PIC)technology is applied to NOMA system in order to solve the problem of high-delay and error propagation in the successive interference cancellation(SIC)algorithm.PIC can effectively compensate for the deficiency of SIC algorithm but its complexity is high.Combined with the advantages of SIC and PIC,a joint interference cancellation(JIC)method is proposed.The error bit performance and algorithm complexity of the three algorithms are simulated and analyzed.The results show that JIC can effectively reduce the delay and improve the error propagation problem in the case of low complexity.Thirdly,in the case of large-scale antennas,the complexity of the the existing zero forcing successive interference cancellation(ZF-SIC)algorithm and the minimum mean square error successive interference cancellation(MMSE-SIC)algorithm is very high.In order to solve this problem,low complexity ND-ZF-SIC algorithm?ND-MMSE-SIC algorithm and ND-ZF-JIC algorithm are proposed.The diagonal matrix decomposition is used to decompose the large matrix into the sum of the diagonal matrix and the hollow matrix.And then,the Neumann series approximation is used to convert the direct inversion of the large matrix into the sum of the products of the diagonal matrix inversion.In order to reduce the complexity while ensuring the accuracy of the approximation,the first two terms of the Neumann series are taken.The simulation results of ZF-SIC?ND-ZF-SIC?MMSE-SIC?ND-MMSE-SIC?ZF-JIC ? ND-ZF-JIC show that the error bit performance of ND-ZF-SIC and ND-MMSE-SIC is similar to ZF-SIC and MMSE-SIC,respectively;The performance of JIC algorithm is better than that of SIC algorithm;ND-ZF-JIC algorithm further reduces the complexity and performance degradation compared with ZF-JIC algorithm.Finally,for the problem that the complexity of the Nuemann series increases rapidly above the second order,JAC-ZF-SIC and JAC-ZF-JIC algorithms are proposed by using Jacobi iteration.The complexity of inverting the matrix is stabilized at the square of K.The Jacobi iterative algorithm is improved by Frobenius matrix decomposition,and the FAC-ZF-SIC and FAC-ZF-JIC algorithms are proposed.The simulation results show that the performance of JAC-ZF-SIC algorithm and JAC-ZF-JIC algorithm are similar to ZF-SIC algorithm and ZF-JIC algorithm respectively.Compared with JAC algorithm,FAC algorithm can converge after two iterations,and performance better than three iterations of the JAC algorithm.