Seismic event and phase detection using time-frequency representation and convolutional neural networks

Past

Abstract: The availability of abundant digital seismic records and successful application of deep learning in pattern recognition and classification problems enable us to achieve a reliable earthquake detection framework. To overcome the limitations and challenges of conventional methods, mainly due to an incomplete set of template waveforms and low signal-to-noise ratio, a generalized model is designed to improve discrimination between earthquake and noise recordings. Using the Synchrosqueezing Wavelet Transform (SWT) of the major seismic arrivals, we train a multi-layer Convolutional Neural Network (ConvNet) to learn general characteristics of background noise and earthquake signals in the time-frequency domain. Deep learning-based methods are superior to traditional techniques in detecting a higher number of seismic events at significantly less computational cost. Bio: Ramin Dokht obtained his Ph.D. in Geophysics from the University of Alberta in 2017. His research was mainly focused on passive imaging of the earth's deep interior and reconstructing long-period seismic data using the spectral characteristics of waveforms. Currently, he is a seismologist in the Geological Survey of Canada, studying the application of deep learning techniques in automatic earthquake detection using the time-frequency analysis of multi-channel data.