Abstract:Seismologists have to deal with overlapping and noisy signals. Techniques such as source separation can be used to solve this problem. Over the past few decades, signal processing techniques used for source separation have advanced significantly for multi‐station settings. But not so many options are available when it comes to single‐station data. Using Machine Learning, we demonstrate the possibility of separating signals for single‐station, one‐component seismic recordings. The technique that we use for seismic signal separation is based on a dual‐path recurrent neural network which is applied directly to the time domain data. Such source separation may find applications in most tasks of seismology, including earthquake analysis, aftershocks, nuclear verification, seismo‐acoustics, and ambient‐noise tomography. We train the network on seismic data from STanford EArthquake Dataset and demonstrate that our approach is (a) capable of denoising seismic data and (b) capable of separating two earthquake signals from one another. In this work, we show that Machine Learning is useful for earthquake‐induced source separation. We provide a reproducible research repository with the algorithms here: https://github.com/crimeacs/source-separation. Earthquake scientists have to deal with overlapping and noisy signals, for example, when an earthquake happens, it often triggers a sequence of aftershocks. Signals recorded from those aftershocks, might overlap significantly. Researches use signal processing techniques to solve this problem. Over the past few decades, these techniques have advanced greatly for multi‐station settings (i.e., when several recorders are available). But the number of techniques able to separate signals on a single‐station data is still limited. Using Machine Learning, we demonstrate the possibility of separating sources for single‐station, one‐component seismic recordings. The technique that we use for seismic signal separation is based on a dual‐path recurrent neural network which is applied directly to the time‐domain data. We present a Deep Learning approach to blind source separation of seismic signals We achieve state‐of‐the‐art performance in denoising and separating earthquake signals Our approach is suitable for early aftershock separation and analysis We present a Deep Learning approach to blind source separation of seismic signals We achieve state‐of‐the‐art performance in denoising and separating earthquake signals Our approach is suitable for early aftershock separation and analysis

Fingerprints of the seismogram’s wavelet analysis results as a tool for creating a compact signal image for the purposes of neural network recognition

Seismic Reflections De-Noising and Recognition Using Empirical Mode Decomposition and Continuous Wavelet Transformation

Automatic Removal of Low Frequency Noise from Seismograms Using the Fingerprinting Method: Algorithms, Properties, Limitations

The Comparison of Time-frequency Analysis Methods and Their Applications

Active Source Seismic Identification and Automatic Picking of the P ̄wave First Arrival Using a Convolutional Neural Network 1

Automatic Waveform-Based Source-Location Imaging Using Deep Learning Extracted Microseismic Signals

Seismic signal recognition by unsupervised machine learning

Interpreting Cross-correlations of One-bit Filtered Seismic Noise

Pattern Recognition Based on Time-Frequency Analysis and Convolutional Neural Networks for Vibrational Events in Φ-Otdr

Feature Extraction of a Non-Stationary Seismic-Acoustic Signal Using a High-Resolution Dyadic Spectrogram

Identification of Seismic Wave First Arrivals from Earthquake Records via Deep Learning.

Seismic Time–Frequency Analysis Via Empirical Wavelet Transform

A machine-learning-based method of detecting and picking the first P -wave arrivals of acoustic emission events in laboratory experiments

Tremor Waveform Denoising and Automatic Location with Neural Network Interpretation

2D Hybrid method:Case of VLF signal amplitude variations in the time vicinity of an earthquake

Efficient similar waveform search using short binary codes obtained through a deep hashing technique

SEDENOSS: SEparating and DENOising Seismic Signals With Dual‐Path Recurrent Neural Network Architecture

&Lt;title>fourier Properties of Fingerprints</title>

A multitask encoder–decoder to separate earthquake and ambient noise signal in seismograms

Heterogeneous Seismic Waves Pattern Recognition in Oil Exploration with Spectrum Imaging