A constrained tensor-based approach for MIMO NL-CDMA systems

In this paper, we propose a deterministic tensor-based approach for joint channel and symbol estimation in the context of multiuser multiantenna (MIMO) CDMA communication systems. We use a new nonlinear (NL) coding allowing to obtain a third-order block-Tucker2 model for the signals received by multiple receive antennas, with a constrained structure for the core tensors that ensures the uniqueness of the tensor model. Two types of receiver are developed. First, assuming that the users' code matrices are mutually orthogonal and known at the receiver, we derive a blind algorithm composed of two steps: a separation of users' contributions in the received signals, with decoding, followed by a blind channel and symbol estimation for each user separately. Then, when the code matrices are unknown, a semi-blind receiver is proposed for jointly estimating the channels, codes and symbols of all the users. Some simulation results are provided to illustrate the performance of the proposed receivers.