Low Complexity ML-Detection of Arbitrary Spherical Codes

Abstract:
We present two algorithms that quantize arbitrary points to a given constellation efficiently, even though there may not be an analytic description of this constellation. We choose spherical codes as constellations, because the best spherical codes can often only be computed numerically. The quantization is equivalent to a maximum likelihood-detection, but the average number of operations can be greatly reduced. The comparison between our proposed algorithms and a brute force approach is done by numerical simulations.