Design of time-frequency distributions for amplitude and IF estimation of multicomponent signals

Abstract

This paper presents a new class of time-frequency. distributions (TFDs) suitable for efficient amplitude and instantaneous frequency (IF) estimation of mono- and multicomponent FM signals. The main characteristic of these TFDs, referred to as T-distributions, is that their kernels are functions of time only in the time-lag domain, hence the name time-only kernels. It is shown that separable time-lag kernels should be lag-independent for best resolution. In addition, amplitude estimation is possible directly from the T-distributions in the case of FM signals, a property that is not verified by other TFDs. The exponential and hyperbolic members of this class are analyzed. Numerical examples are incorporated to show the superior performance of the time-only kernels as compared to their two-dimensional counterparts.