Papers
Topics
Authors
Recent
Gemini 2.5 Flash
Gemini 2.5 Flash
144 tokens/sec
GPT-4o
7 tokens/sec
Gemini 2.5 Pro Pro
46 tokens/sec
o3 Pro
4 tokens/sec
GPT-4.1 Pro
38 tokens/sec
DeepSeek R1 via Azure Pro
28 tokens/sec
2000 character limit reached

Seismic Signal Denoising and Decomposition Using Deep Neural Networks (1811.02695v1)

Published 6 Nov 2018 in physics.geo-ph and eess.SP

Abstract: Denoising and filtering are widely used in routine seismic-data-processing to improve the signal-to-noise ratio (SNR) of recorded signals and by doing so to improve subsequent analyses. In this paper we develop a new denoising/decomposition method, DeepDenoiser, based on a deep neural network. This network is able to learn simultaneously a sparse representation of data in the time-frequency domain and a non-linear function that maps this representation into masks that decompose input data into a signal of interest and noise (defined as any non-seismic signal). We show that DeepDenoiser achieves impressive denoising of seismic signals even when the signal and noise share a common frequency band. Our method properly handles a variety of colored noise and non-earthquake signals. DeepDenoiser can significantly improve the SNR with minimal changes in the waveform shape of interest, even in presence of high noise levels. We demonstrate the effect of our method on improving earthquake detection. There are clear applications of DeepDenoiser to seismic imaging, micro-seismic monitoring, and preprocessing of ambient noise data. We also note that potential applications of our approach are not limited to these applications or even to earthquake data, and that our approach can be adapted to diverse signals and applications in other settings.

Citations (273)

Summary

  • The paper introduces DeepDenoiser, a deep neural network method that significantly enhances seismic signal denoising by learning sparse signal representations.
  • It employs a descending-ascending convolutional architecture with non-linear mapping, achieving notable SNR improvements while preserving waveform integrity.
  • Comparative analysis demonstrates that DeepDenoiser outperforms traditional filtering techniques, with reliable generalization validated on real seismograms from Northern California.

Overview of "Seismic Signal Denoising and Decomposition Using Deep Neural Networks"

The paper "Seismic Signal Denoising and Decomposition Using Deep Neural Networks" by Weiqiang Zhu, S. Mostafa Mousavi, and Gregory C. Beroza introduces a novel method, DeepDenoiser, aimed at improving seismic data processing through deep neural networks. This method addresses the challenge of enhancing the signal-to-noise ratio (SNR) of recorded seismic data, which is crucial for accurate earthquake detection and monitoring.

Methodology

DeepDenoiser is designed to denoise seismic signals by learning a sparse representation of these signals and mapping them into masks to distinguish between the seismic signal and noise. The core network architecture involves a sequence of convolutional layers arranged in a descending-ascending format, facilitating a sparse representation coupled with a non-linear mapping function. The ultimate objective is to apply these masks on the input data to segregate it into signal and noise components effectively.

The training set is composed of synthetic datasets generated by superposing real noise onto clear seismic signals. This supervised learning framework allows the network to learn from both signal and noise data distributions directly. The network's performance is primarily evaluated through metrics such as SNR improvement and waveform shape preservation, benchmarking against traditional methods like spectral filtering and other machine learning-based denoising techniques.

Key Results

DeepDenoiser demonstrates robust performance in denoising seismic signals, even when noise shares the same frequency band as the target signal. The paper highlights an impressive capability to maintain the integrity of seismic waveforms, evidenced by minimal alteration post-denoising. Significant SNR enhancement is achieved across various test scenarios, indicating the method's utility in practical seismic data applications.

Comparative analysis against traditional filtering and general cross-validation (GCV) methods showcases DeepDenoiser's superior efficacy in noise suppression and waveform fidelity. Moreover, application tests on real seismograms from Northern California validates the network's generalizability, further reinforcing its potential for widespread seismic monitoring tasks.

Implications and Future Prospects

The implications of DeepDenoiser extend beyond seismic imaging and earthquake detection to include micro-seismic monitoring and ambient noise preprocessing. Its ability to adapt to different signal types suggests broad applicability across varied domains. Future research directions may focus on enhancing the direct prediction of signal components, potentially overcoming current limitations inherent in the mask-based approach.

DeepDenoiser represents a significant advancement in leveraging deep learning for seismic signal processing. As the field continues to explore more sophisticated neural network architectures and training techniques, the future of seismic data analysis could see further improvements in accuracy and efficiency, thereby advancing earthquake monitoring and related geophysical applications.