In this paper, a two-step procedure is proposed for estimating the parameters of multitone sinusoidal frequency-modulated (MTSFM) signals. In the first stage, coarse parameter estimates are obtained from the instantaneous frequency (IF), while in the second stage, refined estimates are derived from the reconstructed signal phase. Throughout both stages, the multicomponent sinusoidal signal parameters are estimated using fast and robust techniques based on interpolation of the discrete Fourier transform (DFT). Three recent DFT-based estimators are evaluated and compared, with the best performance achieved using the Hybrid AM and q-Shift Estimator (HAQSE). The proposed approach lowers the signal-to-noise ratio (SNR) threshold for resolvable components to below 0 dB. The proposed technique is tested on a human gait micro-Doppler model that allows deviations from the MTSFM model.
Djurović et al. (Wed,) studied this question.