Synchroextracting S Transform and its Application in Seismic Time-Frequency Analysis

Abstract

Summary Non-stationary analysis is always a challenging task because its frequency content changes as time goes. The S transform is widely used to characterize the time-varying features of a non-stationary signal. However, the resolution of S transform subjects to the Heisenberg Uncertain Principle, resulting in energy diffusion in the time-frequency (TF) plane. To overcome this drawback, we develop a new time-frequency analysis method called synchroextracting S transform (SEST). By employing a synchroextracting operator (SEO) in the S transform, the SEST provides a time-frequency distribution with excellent concentration. Differing from the synchrosqueezing transform (SST) by gathering the time-frequency coefficients near the instantaneous frequency (IF) trajectory, the SEST only retains the coefficients most related to the time-varying features of the analyzed signal. Hence, the SEST produces a sparser time-frequency representation (TFR) with a high resolution while allows for signal construction. We validate the proposed method with synthetic examples and compare the result with existing methods. Then, real data examples also illustrate its effectiveness in delineating more distinct channels in the horizon slices.