On the Optimum Design of Space-Time Linear Dispersion Codes

摘要

In this paper, we propose to design linear dispersionrn(LD) codes byminimizing the union bound based on the exact pairwisernerror probability (PEP). We find that the optimization problemisrnnot convex, and we present a convex relaxation that providesrna lower bound on the achievable performance. We employ the gradientrndescent methods to numerically search the optimum dispersionrnmatrices. Simulation results show that codes optimized byrnthe new criterion generally outperform the codes designed basedrnon algebraic number theory.