Time-frequency Analysis Of Time Series Research Articles

A Multi-Taper S-Transform Method for Spectral Estimation of Stationary Processes

The S-transform (ST) is a method of time-frequency analysis of time series. We develop here the Multi-taper S-transform (MTST) for spectral estimation of multi-variate stationary processes through replacing the scalable Gaussian window of the ST with scalable, adjustable orthogonal time-frequency Hermite functions. The MTST is shown to reduce bias and variance of power spectral density (PSD) estimates and coherence over the entire frequency domain and compares favourably with the estimates obtained with the Welch and Multi-taper methods. The MTST method has been successfully applied to data from two anemometers in Hong Kong, during Typhoon Mangkhut, for the estimation of PSD and coherence.

Extraction of forward-directivity velocity pulses using S-Transform-based signal decomposition technique

Robust methods for time-frequency analysis of time series, which provide local information of signals, allow earthquake engineers to study both the input and output of dynamic time history analysis with more reliability. Moreover, time-frequency representations (TFRs) have a major role in the analysis of non-stationary seismic signals exhibiting significant time variation of frequency content. S-Transform (ST) is a modern TFR, which can measure local characteristics of a signal such as amplitude, frequency, and phase at any time instant. This paper presents a new method for decomposition of ground motion signals. A modified version of ST-based technique, originally employed to decompose signals of gearbox vibration, is introduced and applied to the extraction and characterization of pulse-like part of near-fault velocity records, which is contributed to the directivity effects. In addition, a new definition based on ST analysis is used to identify pulse period. The results of implementation of proposed procedure on a database of pulse-like ground motion recordings belonging to the different ranges of magnitude demonstrate the efficiency of proposed method compared with other available approaches. The results, also, indicate that simple approximation of distinct pulses using single-period waveforms, unlike the extracted pulses, cannot represent the impulsive nature of real records adequately.

A procedure based on proper orthogonal decomposition for time-frequency analysis of time series

A procedure for time-frequency analysis of time series is described, which is mainly inspired by singular-spectrum analysis, but it presents some modifications that allow checking the convergence of the results and extracting the detected spectral components through a more efficient technique, especially for real applications. This technique is adaptive, completely data dependent with no a priori assumption and applicable to non-stationary signals. The principal components are extracted from the signals and sorted by their fluctuating energy; moreover, the time variation of their amplitude and frequency is characterized. The technique is first assessed for multi-component computer-generated signals and then applied to experimental velocity signals. The latter are acquired in proximity of the wake generated from a triangular prism placed vertically on a plane, with a vertical edge against the incoming flow. From these experimental signals, three different spectral components, connected to the dynamics of different vorticity structures, are detected, and the time histories of their amplitudes and frequencies are characterized.