Noncoherent Maximum Likelihood Detection for Differential Spatial Multiplexing MIMO Systems

In this paper, we propose a novel noncoherent maximum likelihood detection (NMLD) method for differential spatial multiplexing (SM) multiple-input multiple-output (MIMO) systems. Unlike the conventional maximum likelihood detection (MLD) method which needs the knowledge of channel state information (CSI) at the receiver, WILD method has no need of CSI at either the transmitter or receiver. After repartitioning the observation block of multiple-symbol differential detection (MSDD) and following a decision feedback process, the decision metric of NMLD is derived by reforming that of MSDD. Since the maximum Doppler frequency and noise power are included in the derived decision metric, estimations of both maximum Doppler frequency and noise power are needed at the receiver for NMLD. A fast calculation algorithm (FCA) is applied to reduce the computational complexity of NMLD. The feasibility of the proposed WILD is demonstrated by computer simulations in both slow and fast fading environments. Simulation results show that the proposed NMLD has good bit error rate (BER) performance, approaching that of the conventional coherent MILD with the extension of reference symbols interval. It is also proved that the BER performance is not sensitive to the estimation errors in maximum Doppler frequency and noise power.