Every danger
speaks first.
SonoCop is
built to hear it.
Edge processed
No audio uploaded
Works offline
Book a meeting @ CES 2027, Las Vegas
A threat almost always makes a sound before anything sees it — down an alley no camera covers, in a field at night, on a factory floor nobody is watching. Professional acoustic detection exists, but at prices only large cities can carry.
CCTV cameras per control-room operator in Korea, against a recommended 50.
Ministry of the Interior and Safety, 2025Per square mile, per year, for the incumbent acoustic detection service.
Public procurement recordsTarget price for a complete SonoCop node — array, edge AI and enclosure included.
Ordinary Lab, development targetMultiple units run across the stand, comparing what each one heard. Clap, speak or drop something anywhere in the space, and the readout below shows where it happened and what it was — computed entirely on-device.
Illustration of the bearing readout. Live multi-node demo at our booth.
Eight MEMS microphones in a circular array, equidistant from any incoming sound so nothing arrives distorted.
A lightweight ensemble classifies the sound where it was heard. No audio ever leaves the device.
Neighbouring nodes share what each one heard, turning bearings into a position.
| Specification | SonoCop-8CH-v1 |
|---|---|
| Microphones | 8 × MEMS, circular array |
| Direction finding | ≤ 5°target |
| Sound classes | 20+ at 90%+ accuracytarget |
| Detection to alert | ≤ 10 starget |
| Enclosure | Die-cast aluminium, IP65, −20 to 60 °Cdesign target |
| Power & install | PoE, single cable, 8–10 W, under 5 min |
| Processing | On-device, TPU-expandable — no cloud |
Figures marked target are development goals being validated with KETI. An 8-channel prototype is built and running.
We are raising our next round and looking for robotics and public-safety partners in the US.
Book a meeting