Optimal databased complex demodulation for evolutionary spectra
Kurt S. Riedel, Alexander Sidorenko · IEEE International Conference on Acoustics Speech and Signal Processing · 1993
Complex demodulation of coherent signals in a red noise background is formulated as a kernel smoother in a mixed time-frequency representation. Expected error criteria are used to determine the optimal kernel halfwidth. The authors initially estimate mod delta /sub t//sup 2/A mod /sup 2/ using a higher order kernel smoother with a larger halfwidth to construct a data-adaptive estimate of the evolving amplitude. To estimate the evolving frequency delta /sub t/ phi /sub l/(t), they use a halfwidth corresponding to first derivative estimation.>