Statistical picking of multivariate waveforms

In this paper, we propose a new approach based on fitting a generalized linear regression model to detect change points in the variance of a multivariate covariance Gaussian variable, where the variance function is piecewise constant. Applying this new approach to multivariate waveforms, our method provides simultaneous change point detection on functional time series. The proposed approach can be used as a new picking algorithm to automatically identify the P- and S-waves arrival times on different seismograms recording the same seismic event. A seismogram is a record of the ground motion as a function of time at a measuring station and it typically records motions along three orthogonal axes (x, y, and z), with the z-axis perpendicular to the Earth's surface and the x- and y- axes parallel to the surface, generally oriented in North-South and East-West directions, respectively. The proposed method is tested on a dataset of simulated waveforms, to capture changes in performance to waveform characteristics. Through an application to real seismic data, our results demonstrate the ability of the multivariate algorithm to pick the arrival times in quite noisy waveforms coming from seismic events with low magnitude.