Long before a vessel shows up on radar or a camera feed, its propellers and machinery already fill the water with noise, noise that travels far, works at night, and ignores the weather. That is why underwater acoustics is one of the sensing modalities built into SEAGUARD, the Horizon Europe project developing an integrated approach to maritime cross-border surveillance.
The hard part isn’t hearing the sea. It’s deciding what is being heard.
On 24 September, at SISOM & Acoustics 2026 in the Academy House in Bucharest, our colleagues Teodor Pintilie, Svetlana Segarceanu, and George Suciu presented “Underwater Environment Acoustics Events Classification Using Artificial Intelligence,” work on teaching a neural network to tell cargo ships, tankers, passenger ships, and tugs apart from ambient sea noise using only a few seconds of sound.
The honest result: around 65% accuracy. Ambient noise gets separated almost perfectly, but cargo still gets confused with tankers, and passenger ships with tugs, pairs that are genuinely similar underwater. So the team presented what comes next rather than polishing what they have: distinguishing vessels by size, testing other architectures, and trying gammatone-based spectrograms.
That is exactly why work like this belongs in a room full of people who have spent their careers on acoustic signals. The 37th edition of the Romanian Academy’s symposium, held in its 160th anniversary year, was the right place to be challenged.
Funded by the European Union under Horizon Europe, Grant Agreement No 101168489. Views and opinions expressed are those of the authors only.
Our active project portfolio: MisAIB, ROMEO, PNTS, OPTIX, PROMISE, ENHANCE Europe, WeH 2.0, Wallachia eHUB (WeH), AEROMATICS, NANOEMCA, NetZeroCities, SeaGuard, Museion, SHIELD and SMARDY.
📅 24 September 2026 | Academy House, Bucharest
