A computationally efficient exact ML sphere decoder

Low-complexity tree search based ex act maximum-likelihood (ML) detectors have gained attention for optimum signal detection in multiple-input multiple output (MIMO) wireless systems. In this paper, taking a fresher look at closet lattice point search problem, we propose a new fast exact ML sphere decoder with ever-increasing radius (IR-SD). It is shown that IR-SD visits minimum number of tree nodes and is more computationally efficient than existing sphere decoders. IR-SD is also faster and requires less storage than ML stack detector. An upper-bound of expected computation and storage complexity of IR-SD independent of tree radius is derived. Numerical results validate that IR-SD greatly speeds up the ML detection than other sphere detectors while requiring less storage than ML stack algorithm. A factor-of-2.5 speed-up is observed over the depth-first Schnorr-Euchner sphere decoders when utilizing IR-SD in 16-QAM 10 x 10 complex MIMO system.