A hybrid digital computer-Hopfield neural network CDMA detector for real-time multi-user demodulation

Proposes a hybrid digital computer-neural network multi-user detector whose small computational complexity makes it attractive for real-time CDMA detection. Theoretical results on the nature of the local minima of the optimal multi-user detector (OMD) objective function are summarized, and a method that leads to a significant reduction on the size of the optimization problem to be solved is outlined. The preprocessing problem size reduction stage is followed by a Hopfield neural network employed to solve the irreducible (residual) problem. The performance of the proposed detector is evaluated via simulations and it is shown to exceed that of other suboptimal schemes at a much lower computational cost.<
>