Skip to Main Content
We propose a complex-valued symmetric radial basis function (CV-SRBF) network for nonlinear beamforming in multiple-antenna aided communication systems that employ the complex-valued quadrature phase shift keying modulation scheme. The proposed CV-SRBF classifier explicitly exploits the inherent symmetry property of the underlying data generating mechanism, and this significantly enhances the detection accuracy. An orthogonal forward selection (OFS) algorithm based on the multi-class (four-class) Fisher ratio of class separability measure (FRCSM) is derived for constructing parsimonious CV-SRBF classifiers from noisy training data. Effectiveness of the proposed approach is illustrated using simulation, and the results obtained demonstrate that the sparse CV-SRBF classifier constructed by the multi-class FRCSM-based OFS achieves excellent beamforming detection bit error rate performance.