Reduced-complexity ML detection for coded MIMO systems using an absolute-value search

Spatial multiplexing of data streams on multiple transmit antennae is a well known means of increasing the data-rate in a bandwidth constrained system. At the receiver, maximum likelihood (ML) detection offers the best performance. However, for large constellations and/or large number of antennae the complexity of the decoder grows exponentially. Sphere-decoding is one suboptimal decoding method that has been proposed to reduce complexity by reducing the size of the search space for determining the optimum transmitted symbol vector. In this paper, we present an alternative approach, using an absolute-value based distance for searching, that keeps the size of the search space the same, but reduces the complexity by reducing the number of multiplications required to evaluate the quantities being compared in the search space. This method has vastly reduced complexity compared to ML. Simulation results are presented with a bit-interleaved coded modulation (BICM) system to show that this method has less than 0.5 dB performance loss as compared to ML detection.