High resolution frequency tracking via non-negative time-frequency distributions

A new method for frequency tracking is presented. It combines high resolution time-frequency analysis with the discriminative power of a multiple signal classification (MUSIC, see Schmidt, 1981) technique. The proposed time-varying frequency estimator employs a signal adaptive time-frequency distribution (TFD) with suppressed interference terms. The adaptive TFD is able to resolve components that are closely spaced in time and frequency. The achieved resolution is superior to the one achieved by sliding window techniques which are hampered in joint time-frequency resolution by the uncertainty principle. Simulations show that the presented approach operates reliably in low signal-to-noise ratio environments. The proposed method improves and generalizes the method proposed by Amin and Williams (see IEEE Trans. Signal Proc., vol.46, no.10, pp.2796-2804, Oct. 1998)