Application of the wavelet transform to automatic seismic signal detection

The paper is devoted to such a priority direction of digital seismology as the detection of seismic signals in a noisy medium. Various formulations of the problem of seismic event detection are analyzed. The reasons for the ineffectiveness of the classical formulation of the detection problem are revealed, and an adequate general formulation of the problem in the frequency-time space is proposed. A criterion for the automatic determination of the length of a seismic signal is proposed. The time-scale STA/LTA (Short Time Average to Long Time Average) detector, utilizing the expansion of a seismogram in a packet of wavelets, is proposed. The method eliminates the shortcomings of the standard scheme and enables the detection of seismic events with a low signal-to-noise ratio. The problems of the choice of the mother wavelet and the decomposition level of seismograms are discussed from the standpoint of physical properties of seismic wave fields.