Abstract:Detecting earthquake events from seismic time series has proved itself a challenging task. Manual detection can be expensive and tedious due to the intensive labor and large scale data set. In recent years, automatic detection methods based on machine learning have been developed to improve accuracy and efficiency. However, the accuracy of those methods relies on a sufficient amount of high-quality training data, which itself can be expensive to obtain due to the requirement of domain knowledge and subject matter expertise. This paper is to resolve this dilemma by answering two questions: (1) provided with a limited number of reliable labels, can we use them to generate more synthetic labels; (2) Can we use those synthetic labels to improve the detectability? Among all the existing generative models, the generative adversarial network (GAN) shows its supreme capability in generating high-quality synthetic samples in multiple domains. We designed our model based on GAN. In particular, we studied several different network structures. By comparing the generated results, our GAN-based generative model yields the highest quality. We further combine the dataset with synthetic samples generated by our generative model and show that the detectability of our earthquake classification model is significantly improved than the one trained without augmenting the training set.

What problem does this paper attempt to address?

The problems that this paper attempts to solve mainly focus on two aspects: 1. **Can additional and realistic synthetic data be generated in the case of limited reliable labeled data?** 2. **Can these synthetic data sets be used to further optimize earthquake detection algorithms?** Specifically, the paper aims to solve these problems in the following ways: - **Develop a Generative Adversarial Network (GAN) model**: This model can generate realistic three - dimensional seismic waveform data, including event - type and noise - type. \[ \min_G \max_D V(D, G, x, z)=\mathbb{E}_{x \sim p_{data}}[\log (D(x))]+\mathbb{E}_{z \sim p_z}[\log (1 - D(G(z)))] \] - **Verify the quality of synthetic waveforms**: Verify through visual and quantitative methods, and evaluate using a machine - learning - based earthquake classifier. - **Enhance real earthquake data**: Use the generated synthetic waveforms to expand the actual earthquake data set to improve machine - learning - based earthquake detection methods. ### Background problems Seismic waveform detection is a fundamental task in seismology. However, traditional manual detection methods require a large amount of manpower and are difficult to scale to large - scale data sets. In recent years, automatic detection methods based on machine learning have been developed, but the accuracy of these methods depends on a large amount of high - quality labeled training data. Since obtaining a large amount of labeled data is costly and time - consuming, researchers have explored the possibility of using generative models (such as GAN) to generate synthetic data to alleviate this problem. ### Solutions The author adopts a Conditional Generative Adversarial Network (Conditional GAN), and its value function is shown as follows: \[ \min_G \max_D V(D, G, x, z)=\mathbb{E}_{x \sim p_{data}}[\log (D(x|y))]+\mathbb{E}_{z \sim p_z}[\log (1 - D(G(z|\hat{y})|\hat{y}))] \] where \(y\) is the label of the real sample \(x\), and \(\hat{y}\) is the target label of the synthetic sample. In this way, the generator can generate specific types of seismic waveform data according to the given category. ### Experimental verification To verify the effectiveness of the model, the author designed four experiments: 1. **Visual comparison**: Compare the generated synthetic samples with the baseline model. 2. **Classification task evaluation**: Evaluate the quality of synthetic samples through classification tasks. 3. **Robustness test under a small - scale training set**: Study the performance of the model under limited training data. 4. **Data enhancement application**: Apply the generative model to actual earthquake detection tasks to verify its effectiveness. Through these experiments, the author shows that the generated synthetic data can not only improve the performance of earthquake detection algorithms, but also provide an effective data enhancement means when the labeled data is limited.

SeismoGen: Seismic Waveform Synthesis Using Generative Adversarial Networks

SeismoGen: Seismic Waveform Synthesis Using GAN with Application to Seismic Data Augmentation

Ground Motion Simulation Via Generative Adversarial Network

Augmenting Seismic Data Using Generative Adversarial Network for Low-Cost MEMS Sensors

Data-augmented Landslide Displacement Prediction Using Generative Adversarial Network

SeisGAN: Improving Seismic Image Resolution and Reducing Random Noise Using a Generative Adversarial Network

High Resolution Seismic Waveform Generation using Denoising Diffusion

A Novel Generative Adversarial Network for the Removal of Noise and Baseline Drift in Seismic Signals

STUGAN: An Integrated Swin Transformer-Based Generative Adversarial Networks for Seismic Data Reconstruction and Denoising

Broadband Ground Motion Synthesis by Diffusion Model with Minimal Condition

Deep generative model conditioned by phase picks for synthesizing labeled seismic waveforms with limited data

Generative Adversarial Networks for Labelled Vibration Data Generation

Structural Seismic Response Reconstruction Method Based on Multidomain Feature-Guided Generative Adversarial Neural Networks

High-Quality and Full Bandwidth Seismic Signal Synthesis using Operational GANs

DeepDetect: A Cascaded Region-Based Densely Connected Network for Seismic Event Detection

Semi-Supervised Deep Learning Seismic Impedance Inversion Using Generative Adversarial Networks

Site-Specific Ground-Motion Waveform Generation Using a Conditional Generative Adversarial Network and Generalized Inversion Technique

Semi-Supervised Learning for Seismic Impedance Inversion Using Generative Adversarial Networks

Enhancing Seismic Data Denoising Through Multiscale Analysis Across Transformer and GAN

GAN-LSTM Joint Network Applied to Seismic Array Noise Signal Recognition

Concrete acoustic emission signal augmentation method based on generative adversarial networks