Structure Constrained PARAFAC Model with Application to Signal Processing

Abstract

Structure constrained parallel factor (PARAFAC) model, which combines the structural property of communication signal with the PARAFAC analysis, was presented in this paper. The uniqueness property of the model was discussed and developed. Structure constrained trilinear alternative least square (SCTALS) algorithm was proposed to fit the constrained model. Simulation results showed that, compared to the normal PARAFAC model, the structure constrained PARAFAC was more suitable to describe communication signal. The blind signal processing algorithm based on the constrained model had better fitting performance, and still worked well in some conditions where the normal PARAFAC-based algorithm failed.