Fractional K-best sphere decoding algorithm over rayleigh fading MIMO channels

Abstract

K-best sphere decoding algorithm (KBA) is used to approach near-maximum-likelihood (ML) performance for multiple-input-multiple-output (MIMO) detection with lower complexity than maximum-likelihood (ML) method. In KBA, the value of survivor paths K, can be fixed values only in all tree levels. These fixed values of K's give a certain performances at a certain complexities. In this paper, a new fractional K-best algorithm (FKBA) is proposed which gives a performance and complexity between these discreet performances and complexities for ordinary KBA, acts as if the values of K's are fractions (not integers). This can be achieved by increasing the number of survivor paths into K+Δ in some tree levels and stays K paths in other tree levels. The value of Δ in a specific tree level is resulted from the number of branches have distance metrics lower than or equal the value of average distance metric for all branches in the same tree level. The simulation results show that the performance and complexity of FKBA are approximately in the middle of performances and complexities of two successive values of K (K and K + 1) for different MIMO models of 16 − QAM over Rayleigh fading MIMO Channels for all values of SNR.

Flexible fractional K-best sphere decoding for uncoded MIMO channels

“…The limiting in complexity came from, the big gap between successive complexities of traditional KB [2,3]. Authors in [4] provide a basic solution to this problem by finding only the intermediate complexity level in-between two successive complexity levels, corresponding to two successive values of Ks, based on pruned radius specified in their paper. Beside limitation in complexity problem, many algorithms are presented in [1,2,3,5] to reduce the complexity but without taking the hardware capability into consideration.…”

“…The letter is an extended version of the studies presented in part at an international conference, JEC-ECC 2013 [4].…”

“…The distance metrics dðxÞ ¼ kỹ À Rxk can be computed recursively with partial Euclidean distances (PED) [5]. The upper triangular nature of R allows the optimization problem to be structured as a tree search 2 , with each transmitter representing one level of the tree, and the branches representing a choice of one of the constellation points available for each transmitter [4].…”