Progressive Hypercube Decoding

Abstract

This paper presents a detection technique for multiple-input multiple-output (MIMO) spatial multiplexing systems that is based in a quantized received lattice. This permits the search to be focused into successive ordered subspaces. The proposal greatly reduces the number Euclidian distances need to be calculated replacing it by memory usage for back tracing candidate vectors. Thus, this tool enables further pruning the search tree when using a sphere decoder. The receiver starts by quantizing both the received vector and the lattice points and then defines a neighbourhood around the cell containing the received vector. If needed, the neighbourhood is extended to a larger Manhattan distance. Conversely, dense clusters of lattice points are resolved by increasing the quantization bits per dimension. The paper describes the probability density functions (pdf) of the lattice components and presents numerical results for the pdf of the number of lattice points that are candidates to be evaluated in a subsequent block based on Euclidean distances. All the results are for the frequency flat fast fading channel.