| The phenomenal growth of applications employing machine type communication?MTC?makes MTC become a research focus of the next generation mobile communications.However,a large number of MTC devices pose challenges to traditional random access schemes with their high collision probability and cost of signal overhead.To support the massive connectivity and sporadic traffic requirements,a new random access method based on compressive sensing?CS?is introduced.To improve the accuracy of signal recovery algorithm of this method,improved algorithms for various applications are proposed in this paper.1.To solve the signal recovery problem,the solutions of sub-problems are combined simply in nGOMP algorithm.The relationship among the supports of the solutions in each sub-problem with the same support is ignored,which lead to the low accuracy of signal recovery.To solve this problem,GOMP-LS algorithm is proposed.The support distance function is defined to quantify the difference between supports.Moreover,the optimization model of the shortest distance is modeled to get the optimal support,which is the closest to all sub-problems.Then the problem can be solved by least squares method.In this way the accuracy of signal recovery is improved via the relationship between sub-problems.The simulation results show that the bit error rate?BER?of nGOMP is4.6×10-5,while the BER of the GOMP-LS decreases to1.0×10-5.2.The global optimization support can be calculated after all sub-problems are solved separately,which lead to high computational complicity and latency in GOMP-LS algorithm,which is not suitable for low latency applications.Therefore,the WIGOMP algorithm is proposed for low latency applications.In this algorithm,due to the fact that the supports of solutions of each sub-problem are identical,the weight of each node is changed according to the solution of prior sub-problems to decrease the error probability of the support of posterior sub-problems.In addition,in original algorithm the inverse of the matrix,which just add some new elements,is recalculated in each iteration.To solve the problem of high computational complicity by avoiding several computation of inverse,the inverse Cholesky factorization is employed,and the inverse of the updated matrix is calculated according the matrix factorization in previous iteration.The simulation and computational complicity analysis results show that compared with nGOMP,the BER of proposed algorithm is decreased from4.6×10-5to2.8×10-5,and the computational complicity of nGOMP is 6 times of proposed algorithm.3.To further improve the accuracy of signal recovery of GOMP-LS,which solve each sub-problem separately,WIGOMP-LS algorithm is proposed.In this algorithm the accuracy of signal recovery is improved by increase the weight of nodes corresponding to the elements in support of prior sub-problems,and the errors in the solution of each sub-problem can be further corrected according to the optimization model of the shortest distance modeling and least squares method.The simulation results show that the BER of GOMP-LS is1.0×10-5,while the BER of the WIGOMP-LS decreases to7.4×10-6. |
PDF Full Text Request