降低高条件数信道下的球形译码算法复杂度的方法

A Method of Reducing the Complexity of Sphere Decoder in the High-Condition Number Channel

MIMO系统中,球形译码可以在保证接近ML检测性能的前提下大大降低检测复杂度。但当信道矩阵条件数很高时,球形译码的复杂度仍然会很高。在分析了这一现象的原因后,本文提出了在高层对权值进行合并的一种球形译码算法,因为其减小了译码搜索过程中对树的高层节点的访问的概率,由此降低了搜索复杂度。仿真结果表明,这种算法在低信噪比、高条件数时可以节约20%的浮点运算操作次数。

For MIMO system, as an efficient algorithm to replace ML detection, Sphere Decoding （SD） can reduce the detection complexity greatly. Whereas the bad channel with large condition number still aggravates the computation complexity of SD. In this paper, the reason of this phenomenon is firstly analyzed. Secondly, an algorithm of combining weight at high layer is proposed for SD, aiming to reduce the probability of accessing into the high layer of the tree in the searching process of SD so as to decrease the complexity. The simulation results prove that the proposed algorithm can cut down the complexity distinctly, even to more than 20% at low SNR and high condition number.