Annealed neural network based multiuser detector in code division multiple access communications

Abstract

An annealed neural network based multiuser detector in code division multiple access (CDMA) communications is presented. This detector combines features of the Hopfield neural network and simulated annealing. Unlike existing neural network based detectors, this detector does not need good initial estimates or optimal selection of network parameters. The BER performance of the proposed detector is close to the theoretical lower bound of the BER peformance, especially in ‘near–far’ situations. Extensive numerical evaluation of the proposed technique as well as various suboptimal multiuser detectors is conducted using Monte-Carlo simulation.