A Hybrid MMSE and K-Best Detection Scheme for MIMO Systems

Abstract

A new multiple-input multiple-output (MIMO) detection scheme combining minimum-mean-square-error (MMSE) detection and the K-best detection algorithm is proposed. The proposed scheme leverages the MMSE detection results to ease the demand of a large K in the conventional K-best algorithm to achieve satisfactory performance. The post-detection SNR obtained after MMSE detection is consulted to determine the symbols upon which a reduced-dimension K-best algorithm (hbest algorithm) is performed to obtain final detection results. Parameters associated with the proposed scheme are empirically chosen to make a fair comparison with the conventional K-best algorithm. Extensive Monte Carlo simulation demonstrates that the hybrid approach exhibits significant performance gain over both MMSE and K-best detection schemes.Index Terms-MIMO systems, maximum likelihood detection, MMSE detection, K-best detection algorithm.

A partial inter-layer parallel sphere decoder for multiple-input multiple-output systems

“…On the other hand, although existing breadth-first algorithms such as K-best have been realized in hardware implementations with parallel data paths [5], they have only explored data parallelism within each layer in the search tree, inter-layer search process is still intrinsically serial. Some recent approaches, such as hybrid MMSE and K-best detection [6], bounded selective spanning with extended fast enumeration [7], also fail in exploring inter-layer parallelism. This intrinsic serial property will result in very ineffective resource utilizations and limit the peak detection speed on parallel architectures.…”