Adaptive order selection in quasi-maximum likelihood-based IF estimation

The paper addresses adaptive order selection in a recently proposed quasi-maximum likelihood (QML) method for the instantaneous frequency (IF) estimation, which models the considered signal by a polynomial-phase signal (PPS). However, when the PPS order is not known in advance, or when signal with a non-polynomial IF is considered, the QML method is repeated for polynomial orders higher than the optimal one. We propose to perform the QML-based IF estimation for successive increasing values of polynomial IF order of the modeling PPS until a sufficiently accurate IF approximation of the considered signal is obtained. In this way, the estimation procedure is not repeated for polynomial orders higher than the optimal one, thus significantly reducing the calculation complexity.