Parameter estimation for HFM signals using combined STFT and iteratively reweighted least squares linear fitting

• A method combing STFT and IRLS linear fitting for estimating parameters of HFM signals is proposed.
• The proposed method transforms a complex nonlinear problem into a linear one.
• The lower bound of the estimator's variance and bias of the parameters of HFM signals are derived.
• Both simulation results and processing of sea trial data are presented.
• The proposed method is generic and may be applied to any TF representation.

This paper presents a fast and robust method for estimating the starting frequency and period slope of hyperbolic frequency modulated (HFM) signals. The method involves, first, the instantaneous frequency (IF) estimation of HFM signals based on the peak of short-time Fourier transform (STFT) and, second, taking reciprocal of the estimated IF to get the zero crossing interval (ZCI). Parameter estimation of HFM signals is then achieved by using iteratively reweighted least squares (IRLS) linear fitting method to fit the ZCI which is a linear function of time. Both the approximate analysis of the magnitude spectrum and the formula used to determine the window length of STFT are derived for HFM signals. The lower bound of the estimator's variance and bias of the parameters of HFM signals are also derived in order to compare the performance of the proposed method. At last, both the simulation results and processing of sea trial data are presented to justify the validity and feasibility of the proposed method.