The use of sliding spectral windows for parameter estimation of decaying sinusoidal signals

It is quite common to model speech as the sum of a number of exponentially damped sinusoids. An example of where this is important is in the estimation and tracking of the formants of the speech. Typically, parametric spectrum analysis methods are used to estimate the parameters of the decaying sinusoids (or modes) Prony's method, Burg's method, and Pisarenko's method are some of the methods employed. This paper will present a Fourier based algorithm for estimating the parameters of damped sinusoidal components. The algorithm is based on the sliding window method of Poon and Lee (1988), but has a number of innovations.