Optimal short-time Fourier transform for monocomponent signals

Abstract

New methods of improving the short-time Fourier transform representation of signals have recently emerged. These methods use linear canonical transforms to bring the signal into a minimal time-bandwidth product form. Here we show that linear canonical transforms are not sufficient to achieve the minimum time-bandwidth product for high-order modulated mono-component signals. Therefore we propose a novel short-time Fourier transform method which requires an adaptive window, making use of an initial instantaneous frequency estimator. The new approach is able to achieve the highest possible resolution for monocomponent signals. Finally, we discuss the benefits of the proposed method. © 2004 IEEE.