High-resolution data-adaptive time-frequency analysis

Capon's filters were originally proposed in array signal processing. It had been shown that they can be an efficient spectral analysis tool. Assuming that the signal frequency is constant the Capon's version of the short-time Fourier transform is derived in time-frequency analysis. In this paper we have generalized this approach to the cases when the instantaneous frequency is not constant. The Capon's forms of the Wigner distribution (WD) and the local polynomial periodogram are introduced. Performance improvement of these representations over the standard ones is illustrated by examples.