Analysis of multicomponent signals by multilinear time-frequency representations

The aim of this work is the analysis of a method for the detection and parameter estimation of polynomial-phase, mono or multicomponent, signals embedded in white Gaussian noise, based on multilinear time-frequency signal representations. The proposed approach, based on a proper coherent integration of the multilinear time-frequency representation along paths depending on the model assumed for the instantaneous phase of the useful signal, presents some advantages with respect to conventional techniques, based on multilinear time-frequency transforms, in terms of: (i) a closer approach to the Cramer-Rao lower bounds, (ii) a higher output signal-to-noise ratio, and (iii) a better capability of discriminating multicomponent signals