Generalized PARAFAC model for multidimensional wireless communications with application to blind multiuser equalization

In this work we develop a new tensor modeling approach for multinser wireless communication systems where the received signal has a multidimensional nature. The proposed tensor model follows from a third-order (3D) Block-Parallel Factor (Block-PARAFAC) decomposition with factor interactions, which can be viewed as a more general model than the standard model [1], [2]. The proposed tensor decomposition aims at unifying the received signal modeling for i) Temporally-Oversampled, ii) Direct-Sequence Code Division Multiple Access (DS-CDMA) and iii) Orthogonal Frequency Division Multiplexing (OFDM) systems. This modeling approach assumes a receiver antenna array, specular multipath propagation and frequency-selectivity. We show that the model for each of the considered systems can be derived from the Block-PARAFAC model by making appropriate choices in its dimensions and/or structure. As an application of the proposed tensor model to blind multiuser separation/equalization, a new receiver algorithm is derived.