Interference Cancellation Based Detection for V-BLAST With Diversity Maximizing Channel Partition

Multiple-input multiple-output (MIMO) systems achieve very high bandwidth efficiencies through spatial multiplexing. However, the complexity of optimal detection in such systems motivates the need for more practical alternatives. Recently, a suboptimal lower complexity detection scheme called ¿generalized parallel interference cancellation¿ (GPIC), with close to optimal performance, was introduced. The reported performance of GPIC, however, was assessed by computer simulations only. In this paper, we show that with its original design, GPIC does not always provide close to optimal performance. Based on a diversity analysis of GPIC like techniques, we propose two new improved algorithms, referred to as Sel-MMSE and Sel-MMSE-OSIC, and derive sufficient conditions for achieving optimal performance asymptotically. We also provide a complexity analysis of these two schemes, and show that for large constellation sizes it is lower than the original GPIC. While still more complex than the fixed complexity sphere decoder by a factor in the range of 2-3 (for most configurations), our algorithms are also applicable to undetermined MIMO systems. Simulations results confirm that the new schemes provide maximal diversity gains. Furthermore, Sel-MMSE-OSIC provides a significant gain over Sel-MMSE, making its performance nearly indistinguishable from optimal for all signal-to-noise ratio (SNR) levels.