A Deterministic Method for Choosing Search Radii in Sphere Decoding

Sphere decoding is a widely used technique in communications. Given a received noisy signal, this method retrieves the source signal by exhaustively searching for an optimal solution in a hypersphere. Apparently, choosing an appropriate radius for a search sphere has significant impact on the complexity of sphere decoding. A too large sphere requires prohibitive cost of searching, while a too small sphere contains no solution. In this paper, we first describe a radius selection method which produces tight search sphere. However, due to inexact floating-point computation, the computed radii may be too small for the search spheres to contain solutions. We then perform an error analysis and propose a modified radius selection algorithm by incorporating rounding errors. Finally, we demonstrate our experiment results.