Original title: Biometric verification using oral acoustic reflections
Amit Kam, Nir Iakoby, Daniel Tubiana, Eran Hochstadter +2 · September 2026
Common identity checks like fingerprints or facial scans have real downsides in cost, accuracy, delay, and faking risks. To tackle this, researchers created Oral Cavity Acoustic Verification, or OCAV. It checks who you are using the inside of your mouth. The tool works by sending sound waves directly into your mouth. It then listens to the acoustic response—which is how those sound waves bounce back. Next, it compares those reflections to a saved acoustic signature taken from you earlier. If they match, it verifies your identity. The team found that OCAV reliably tells people apart, and combining a few quick tests in a row improves its accuracy even further. Because it actively sends its own sound into the mouth instead of listening to you speak, it may resist fake audio attacks better than regular voice checks. It is designed for tools that already touch the mouth, like medical inhalers.
You have a set of keys and need to pick the right one in the dark. Feeling each key's shape would work, but a recording of their jingle wouldn't help because you need the actual physical key. Here, the system sends a sound into your mouth and listens to the echo created by your mouth's unique shape, rather than relying on a sound you produce yourself. This makes it much harder for an imposter to trick the system with a recording.
This means your inhaler or breathalyzer could check who you are before delivering a dose or allowing ignition. Unlike voiceprints, it resists recording-based fakes because it actively probes your mouth's shape. It could help enforce age restrictions on vaping or catch drunk drivers before they start the car.
The testing was done in a lab with 108 healthy volunteers aged 18-60; real-world accuracy on more diverse populations is unknown. The system requires direct oral contact, so it cannot be used for remote verification. Accuracy is 90.1% for a single trial, meaning about one in ten attempts would wrongly accept or reject someone; higher confidence requires pooling five trials. The abstract does not discuss speed, cost, or performance in noisy environments.
Ask anything about this paper. Answers come from its abstract, and say so when the abstract doesn't cover your question.
Biometric verification using oral acoustic reflections — Amit Kam, Nir Iakoby, Daniel Tubiana et al. — PLoS ONE — doi:10.1371/journal.pone.0359160
Abstract licence: CC BY
Explanation generated by Synthesis from the abstract — it is not the original paper.