13-09-2012, 05:27 PM
Time-Frequency Analysis of Non-Stationary Signals Using Improved S-Transform
Time-Frequency Analysis.pdf (Size: 494.93 KB / Downloads: 60)
ABSTRACT:
Spectral analysis using the Fourier Transform is a powerful technique for stationary time series where
the characteristics of the signal do not change with time. For non-stationary time series, the spectral content
changes with time and hence time-averaged amplitude spectrum found by using Fourier Transform is inadequate
to track the changes in the signal magnitude, frequency or phase. This paper presents a new time-frequency
signal analysis method, called Improved S-Transform (IST) for visual localization, detection of various nonstationary
power signals. Among various windows Gaussian window is found to provide excellent normalized
frequency contours of the power signal disturbances, suitable for accurate detection, localization. This paper
presents time-frequency contours of various non-stationary power quality signals are processed through the
Improved S-Transform (IST) with Gaussian window.
INTRODUCTION
Time–frequency analysis is has been successfully
used in dealing with rapidly varying transient signals,
such as guided- wave signals and damping vibration
signals [1]. For Time–Frequency Representations
(TFRs), the Short-Time Fourier Transform (STFT), the
Wigner-Ville distribution (WVD) and the Wavelet
Transform (WT) are commonly used. STFT and WVD
have certain advantages over the WT, but they also have
some critical limitations in comparison with the WT. The
fixed time–frequency window of STFT can lead to
undesirable time and frequency resolutions. In spite of its
excellent time–frequency resolution, using WVD, it is
often difficult to analyze a signal with compositefrequency
components because of the appearance of
interference terms.
APPLICATIONS TO PQ DISTURBANCES
AND SIMULATION RESULTS
In our study we have discussed different types of power
signal waveforms such as voltage swell, flicker and
momentary interruption. The time-frequency contours of
these disturbances and their corresponding change in
magnitude vs. time are analyzed with MATLAB
software. The chosen sampling rate is 3.84 kHz. The IST
outputs show the plots of the normalized frequency
contours of a given magnitude in the time-frequency coordinate
system.
CONCLUSION
This paper has proposed a new approach for
detection and localization of power quality
disturbances in a power distribution system. The
Improved S-Transform (IST) with a variable window
(Gaussian window) as a function of PQ signal
frequency is used to generate contours. In future
this method can be widely applied to image
processing, Radar signal detection and
classification and seismic signal processing.
Automatic classification can be done by extracting
feature vector from the IST frequency contour and
finally passing those pertinent feature vectors
through an intelligent classifier for pattern
classification.