Improved instantneous frequency estimation using an adaptive short-time fourier transform

Instantaneous frequency estimation (IFE) arises in a variety of important applications, including FM demodulation. We present here a new time-frequency representation (TFR)-based approach to IFE based on an adaptive short-time Fourier transform (ASTFT). This TFR leads naturally to a type of short-term ML estimator of the IFE. To further improve the performance, we apply a hidden Markov model (HMM)-based post-estimation tracker. The end result is an up to 12-dB reduction in the threshold SNR over frequency discrimination, an 8-dB improvement over the phase-locked loop (PLL) for a Rayleigh fading channel, and the best low-SNR performance of any known IF estimator for FM demodulation.