S. Ezekiel, M.J. Barrick, and M. Lang (USA)
Seismic Signal, Fractal dimension, Correlation dimension, Multifractal Spectrum
Multifractal analysis is a new and promising approach to analyze and characterize non-stationary signals such as seismic signals, electrocardiograms, heart rate variability signals, stock prices, etc. Traditional Fourier analysis imposes the assumption that signals are stationary over temporal segments. Such an assumption is inappropriate for seismic signals because conditions in mines undergo constant change. Multifractal analysis does not impose this assumption, and is therefore useful for processing seismic signals. In this paper we propose a novel correlation dimension based method to analyze seismic signals, which is superior to many conventional used methods. The correlation fractal dimension is calculated for overlapping windows for the entire signal producing a multifractal spectrum. Using this strategy our method pinpoints seismic events in the seismic signal. Our results show that a small correlation dimension corresponds to a region of seismic activity in the signal. The approach applied here may be useful for analyze of other non stationary signals.
Important Links:
Go Back
