![[2602.20317] Fast Spectrogram Event Extraction via Offline Self-Supervised Learning: From Fusion Diagnostics to Bioacoustics](https://arxiv.org/static/browse/0.3.4/images/arxiv-logo-fb.png){kind=logo}
[2602.20317] Fast Spectrogram Event Extraction via Offline Self-Supervised Learning: From Fusion Diagnostics to Bioacoustics
Summary
This article presents a self-supervised learning framework for the automated extraction of coherent and transient modes from high-noise time-frequency data, addressing challenges in analyzing large datasets from fusion diagnostics and bioacoustics.
Why It Matters
As next-generation fusion facilities generate massive amounts of data, efficient analysis methods are crucial. This framework not only enhances data processing speed but also supports real-time mode identification, which is vital for advanced plasma control and other applications in signal processing.
Key Takeaways
- Introduces a self-supervised learning framework for signal extraction.
- Addresses data analysis challenges in fusion diagnostics and bioacoustics.
- Achieves real-time mode identification with an inference latency of 0.5 seconds.
- Utilizes non-linear optimal techniques in multichannel signal processing.
- Demonstrates applicability across various diagnostic measurements.
Electrical Engineering and Systems Science > Signal Processing arXiv:2602.20317 (eess) [Submitted on 23 Feb 2026] Title:Fast Spectrogram Event Extraction via Offline Self-Supervised Learning: From Fusion Diagnostics to Bioacoustics Authors:Nathaniel Chen, Kouroche Bouchiat, Peter Steiner, Andrew Rothstein, David Smith, Max Austin, Mike van Zeeland, Azarakhsh Jalalvand, Egemen Kolemen View a PDF of the paper titled Fast Spectrogram Event Extraction via Offline Self-Supervised Learning: From Fusion Diagnostics to Bioacoustics, by Nathaniel Chen and 8 other authors View PDF HTML (experimental) Abstract:Next-generation fusion facilities like ITER face a "data deluge," generating petabytes of multi-diagnostic signals daily that challenge manual analysis. We present a "signals-first" self-supervised framework for the automated extraction of coherent and transient modes from high-noise time-frequency data. We also develop a general-purpose method and tool for extracting coherent, quasi-coherent, and transient modes for fluctuation measurements in tokamaks by employing non-linear optimal techniques in multichannel signal processing with a fast neural network surrogate on fast magnetics, electron cyclotron emission, CO2 interferometers, and beam emission spectroscopy measurements from DIII-D. Results are tested on data from DIII-D, TJ-II, and non-fusion spectrograms. With an inference latency of 0.5 seconds, this framework enables real-time mode identification and large-scale autom...
