Constrained detection for multiple-input multiple-output channels

We develop a family of constrained detectors for multiple-input multiple-output (MIMO) channels by relaxing the maximum likelihood (ML) detection problem. Real constrained linear detectors and decision feedback detectors are proposed for real constellations by forcing the relaxed solution to be real. Generalized minimum mean-square error and constrained least squares detectors are generalized as MIMO detectors for both constant and non-constant modulus constellations. Using our constrained linear detectors, we propose a new ordering scheme to achieve a tradeoff between interference suppression and noise enhancement. Moreover, we introduce a combined constrained linear and decision feedback detector to mitigate the error propagation in decision feedback. Simulation results show that the combined detectors achieve significant performance gain over V-BLAST detection.