Adaptive multiuser detection in DS/CDMA systems using generalized regression neural network

Artificial neural networks are extremely used for detection of spread-spectrum signals in multiple-access environments. In this paper we suggest the use of generalized regression neural networks (GRNN) on multiuser detectors in DS/CDMA systems. The network is trained by applying the estimated joint probability density function. After training, the network can obtain the required timing without knowing the signature waveforms and the received signal amplitudes. The simulation results demonstrate that the proposed receiver has higher performance in comparison to detectors which have more knowledge of system parameters.