Subspace Hebbian Learning and Maximum Likelihood ICA Based Algorithms for Blind Adaptive Multiuser Detectors

In this paper, two independent component analysis (ICA) based algorithms are proposed for blind adaptive multiuser detection (MUD) in CDMA systems. The first algorithm is subspace Hebbian learning and the second one is subspace maximum likelihood (ML). Signal subspace estimation is employed for data whitening as a preprocessing step in both algorithms. The performances of the algorithms are evaluated using computer simulations. Simulation results show that the subspace Hebbian learning algorithm converges faster in the expense of a little inferiority in the steady state error compared to that of the subspace ML algorithm. The steady state performances of the two algorithms are also compared to that of minimum output energy (MOE) detector.