Metric-segmented low-complexity ML detection for spectrum-efficient multiple-antenna systems

In this paper, we reveal that a maximum-likelihood detection (MLD) for multiple-input multiple-output (MIMO) systems requires not exponential complexity order in multiplication operations but cubic order against the number of antennas, by introducing the so-called correlation metric. However, the MIMO-MLD requires still exponential order in arithmetic additions. In order to further reduce the complexity, we propose effective techniques referred to as a metric-segmentation and a norm-constraint approach. We show that the proposed MLD using these techniques can significantly reduce the computational complexity without any performance degradation. Moreover, the proposal does not need additional high-complexity operations such as QR decompositions and matrix inversions.