Combined PARAFAC-Subspace Approach to Blind Multiuser Equalization

This paper presents a tensor decomposition known as PARAFAC (Parallel Factors) and propose a new blind multiuser equalization approach for wireless communication systems, employing an antenna array and oversampling at the receiver. First, a tridimensional PARAFAC approach for modeling the received signal is proposed, the 3 dimensions being space, time and oversampling dimensions. Then, a blind multiuser receiver performing multiuser signal separation and equalization is formulated, combining PARAFAC modeling and a subspace method. The key difference of the proposed approach compared to most of existing ones is on the fact that the inherent tensor structure of the received signal is exploited. The proposed PARAFAC receiver has two blind-processing stages. In the first one, co-channel user signals are separated in the tensorial domain using an alternating least-squares algorithm. In the second stage, a subspace method is used to independently equalize each user sequence. Simulation results are provided to illustrate the performance of the proposed receiver. Our results show that the PARAFAC receiver performs closely to the MMSE (Minimum Mean Square Error) and ZF (Zero Forcing) receivers.