4×4 V-BLAST系统分组最大似然检测算法

Group Maximum Likelihood Detection Algorithm for a 4×4 V-BLAST System

对于V-BLAST系统的检测,最大似然(ML)算法有着最优的性能却也有最大的计算复杂度;经典的排序连续干扰抵消(OSIC)算法复杂度较低,但数值稳定性差,且性能与ML差距较大。因此,本文基于检测性能和计算复杂度折中的思想,针对4×4 V-BLAST系统提出了一种分组最大似然(Group ML,GML)检测算法,在保证较好检测性能的基础上,通过将四维ML检测器分成两组二维ML检测器来降低计算复杂度。此外,本文还提出了一种简化的最大似然(Simpli-fied ML,SML)检测算法,通过将每组中的二维ML检测器的搜索空间从二维降至一维,进一步降低了计算复杂度,并证明其与ML算法具有一致的性能。仿真表明,在误符号率为10-3时GML算法相比OSIC算法有约7dB的性能提升。经分析知,GML算法复杂度与ML-OSIC算法相比在高阶调制方式下有着显著的降低,易于硬件实现。

Maximum likelihood（ML） algorithm not only has optimal performance but also has maximal computational complexity for detection of V-BLAST systems. Classical ordered successive interference cancellation（OSIC） algorithm has much lower complexity. However, it has a numerical value stability problem and has a largish performance gap by comparison to ML algorithm. Therefore, a group maximum likelihood （GML） algorithm is proposed for a 4×4 V-BLAST system in this paper, which is based on tradeoff between detection performance and computational complexity. On the basis of guarantee of preferable detection performance, it divides a four-dimension （4-D） ML detector into two 2-D ML detectors to reduce computational complexity. Moreover, a simplified maximum likelihood （SML） algorithm is also proposed to reduce further complexity by reducing the searching space of each 2-D ML detector from 2-D to 1 - D. Simulations show that GML algorithm has about 7dB performance gain compared with OSIC algorithm at symbol error rate equals to 10 -3. After analysis, GML algorithm＇ s complexity reduces remarkably in large constellation modulation manner in comparison with that of ML-OSIC algorithm. This facilitates the hardware implementation.