Euclidean Distances in Quantized Spaces with Pre-stored Components for MIMO Detection

This paper proposes a technique to reduce the complexity involved in the maximum likelihood detection of multiple input multiple output (MIMO) spatial multiplexing systems. Both the received lattice and the components of the received signal are quantized, corresponding to a mapping into a multidimensional space divided into hypercubes. In the new space the maximum likelihood detection criterion can be applied making use of a small look-up table storing all the exact possible distance components in each dimension of the quantized space. The number of pre-stored elements can be as small as the number of quantization levels per dimension. This procedure eliminates the multiplications involved in the calculation of the squared Euclidean distances. The impact of the quantization error is assessed by means of simulations over fast flat fading channels. Near optimum performance is achieved with only 5 bits representing each dimension of the received signal.