Multi-User Detection Based on Improved KICA with Bat Algorithm

Abstract

In this paper, an improved kernel independent component analysis (KICA) algorithm is proposed for multi-user detection (MUD). In this algorithm, a new hybrid kernel function is adopted. In addition, the bat algorithm is applied to the optimizing process of independent component separation. Simulation results show that the new hybrid kernel function performs better in MUD than other kernel functions, and the improved KICA with bat algorithm has the smallest bit error rate (BER) when compared with classical FastICA and KICA algorithms.