Solar filaments, giant strands of scorching gas suspended in magnetic fields whose lines can reconnect during flares, vibrate like the strings of an unseen instrument.

Scientists used to search for these oscillations manually, poring over images, which took a lot of time. Now an automated 'listener' based on machine learning has been built. It analyzes solar images in the spectral line of hydrogen Hα (discovered by Johann Balmer), applying a mathematical technique rooted in the ideas of Joseph Fourier to detect periodic changes in brightness. To avoid confusing real vibrations with noise, the program cross-checks with computer simulations.

In two weeks, the robot caught 91 oscillations compared to 22 by a human — much like Shazam recognizes a song in a noisy crowd. From the rhythm of the tremors and their decay rate, one can calculate the strength of the magnetic field anchoring the filament, akin to determining a string’s tension by listening to its sound. The slowest vibrations, lasting hours, went unnoticed by the human eye, but the program captures them too. This provides deeper insight into the Sun's magnetic processes and allows more precise prediction of solar activity. This approach opens a new era in observing our star.