Deterministic exponential modeling techniques for high spectral resolution time-frequency distributions.
Spectral resolution of t-f distributions can be improved by modeling the local autocarrelation function (LAF) by stochastic or deterministic models. This requires new sets of constraints to be imposed on the t-f kemels. One set of constraints must allow the local m o w band components in the data to be presented by damped/undamped sinusoidal components in the LAF. Another set of constraints must be considered for noisy environment to maintain the uncorrelatedness between the signal and the noise components in the LAF. This paper presents the class of kernels satisfying the first set of constraints leading to exponential deterministic modeling of the LAF. The effects of the existing desired kemel properties on the noise characteristics are also delineated.
|Main Author:||Amin, Moeness G.|
|Other Authors:||Williams, William J.|