Abstract:

It was assumed that seismic signals follow Gaussian distribution. After analyzing the dynamic sample variance of real seismic data, the authors proposed that seismic data obeys nonGaussian α-stable distribution. In the applications of non-Gaussian α-stable distribution seismic signal processing, such as noise suppressing and seismic impedance inversion, characteristic exponent is a key parameter; And if the norm greater than characteristic exponent,  misleading results will be produced. Combining with particle swarm optimization algorithm,  α-stable distribution characteristic exponent estimation method based on fractional lower order moments was applied to real seismic data and the characteristic exponent of real seismic data was obtained. The work of this paper shows that the pulse characteristic in real seismic data are stronger than that in Gaussian distribution; The proposition that real seismic data follow non-Gaussian α-stable distribution is reasonable; The calculation of real seismic data character exponent is also helpful to the selection of seismic signal processing algorithm.

Key words: non-Gaussian, α-stable distribution, characteristic exponent, signal processing