SoundShield

Important results from the project

SoundShield advanced from an early prototype toward a more integrated AI-based acoustic surveillance system. About 30,000 audio clips were collected in field tests, and well-shielded units detected tested drones at about 0.8-1.2 km. New custom hardware, improved data processing and a mobile backend were developed. Generative AI also showed promising few-shot adaptation. Full TRL 7 is not formally verified in the available evidence.

Expected long term effects

In the longer term, the results may provide a scalable and cost-effective complement to camera and radar for military and civil security. Field data and hardware lessons support more robust product development, while few-shot AI may enable faster adaptation to new sounds and threats. The backend and mobile functionality provide a foundation for customer trials, further verification and future dual-use commercialization.

Approach and implementation

The project was carried out iteratively across AI, hardware, data collection and infrastructure. Two Tibble field tests produced about 30,000 audio clips and showed strong drone detection plus electrical-noise issues. The first hardware batch required troubleshooting and a second revision was produced. AI work covered improved deep learning and generative/few-shot methods, while backend and mobile functions were developed. Findings shifted focus toward shielding, data quality and robustness.