GNGTS 2018 - 37° Convegno Nazionale

758 GNGTS 2018 S essione 3.3 Oropeza, V., and M. Sacchi; 2011: Simultaneous seismic data denoising and reconstruction via multichannel singular spectrum analysis , Geophysics, 76, V25–V32. Pathak, D, et al.; 2016: Context encoders: Feature learning by inpainting , in Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition, pp. 2536–2544. Ronneberger, O. et al.; 2015: U-net: Convolutional networks for biomedical image segmentation , International Conference on Medical image computing and computer-assisted intervention, Springer, 234–241. Siahkoohi, A. et al.; 2018: Seismic data reconstruction with generative adversarial networks , in 80th EAGE Conference and Exhibition 2018. Spitz, S.; 1991: Seismic trace interpolation in the fx domain , Geophysics, 56, 785–794. Xu, S. et al.; 2005: Antileakage fourier transform for seismic data regularization , Geophysics, 70, V87–V95. Wang, B. et al.; 2014: Dreamlet-based interpolation using pocs method , Journal of Applied Geophysics, 109, 256– 265. AUTOMATIC DISCRIMINATION AND FAST WAVEFIELD DECOMPOSITION OF TECTONIC ANDVOLCANO-TECTONIC (VT) EARTHQUAKES BY INDEPENDENT COMPONENT ANALYSIS S. Petrosino 1 , E. De Lauro 2 , M. Falanga 3 1 Istituto Nazionale di Geofisica e Vulcanologia - Sezione di Napoli, Osservatorio Vesuviano, Napoli, Italy 2 Università di Roma Tre - Dipartimento di Architettura, Roma, Italy 3 Università degli Studi di Salerno, Dipartimento di Ingegneria dell’Informazione ed Elettrica e Matematica applicata/DIEM, Fisciano (SA), Italy The discrimination of seismic events from the ambient background noise, the wavefield decomposition into basic wave-packets and the identification of the main phases are fundamental tasks in the study of the earthquake source processes. In the last years, the Independent Component Analysis (ICA), a technique used in advanced signal processing to separate statistically independent sources (Hyvärinen et al. , 2001), has found relevant and interesting application in the seismological field both for event detection and discrimination purposes (Ciaramella et al. , 2011, Capuano et al. , 2017), as well in providing spectral decomposition of the wavefield into basic wave-packets related to the source and their polarization pattern (Capuano et al. , 2016). Recently, De Lauro et al. (2016) have shown the efficiency of the ICA in decomposing seismic wavefield of volcano-tectonic (VT) earthquakes recorded at Campi Flegrei by the local network into basic wave-packets naturally polarized in the vertical and horizontal planes, providing a clear identification and separation of the P and S waveforms in time domain. The ICAperforms a decomposition of the signal mixtures into independent time sources. The mixing model is written as, where x is an observed m -dimensional vector (the seismogram), s is an n -dimensional random vector whose components are assumed to be mutually independent; a ij are the constant elements of an unknown m x n mixing matrix A . The extension to the frequency domain for sources that are convolved (CICA) gives the model: where k is a suitable time lag. The number of extracted sources s (Independent Components, ICs) for each direction of motion North-South, East-West and Vertical (NS, EW and Z) is at least equal to the number of the input waveforms, so it depends on the number of seismic stations. De Lauro et al. (2016) modified the basic technique in order to apply it to the three directions of motion: the NS, EW and Z components at each station are treated as instantaneous mixtures