Robust blind multiuser detection based on the worst-case performance optimization of the MMSE receiver

The conventional minimum mean-square error (MMSE) multiuser receiver requires that each particular user transmits a sequence of training symbols known to the receiver and the receiver estimates the user signatures using this knowledge. However, in the presence of multiaccess interference (MAI) during the training period and/or in scenarios with short training sequence length, the signature estimates can be erroneous, and the performance of the MMSE multiuser receiver can degrade substantially. In this paper, we propose a new blind multiuser receiver that is robust against the effects of erroneously presumed desired user signature and short data length. Our approach is based on the explicit modeling of possible mismatches in the mean-square error cost function and worst-case performance optimization. We show that this approach leads to a multiuser receiver which uses the data covariance matrix with an adaptive diagonal loading. The proposed method has simple implementation with a computational complexity comparable with that of the fixed diagonal loading-based multiuser receiver. Simulation results show performance improvements achieved by our approach relative to existing techniques.