Averaging methods for blind multiuser detection in DS/CDMA systems

We present a blind adaptive code-aided algorithm for suppression of multiuser interference (MAI) and narrow-band interference (NBI) in DS/CDMA systems. The algorithm -which we call the BAG algorithm-is based on the concept of accelerating the convergence of stochastic gradient methods by averaging. This ingenious concept of averaging was invented by Polyak-our paper examines its application to interference suppression in DS/CDMA systems. We prove that our averaged gradient algorithm has the same asymptotic rate of convergence as RLS, but at a order of magnitude lower computational cost.