01
Published task benchmark
94.29% for 1-second snore-segment classification and 88.49% for 200-second breathing-window classification, evaluated against PSG reference annotations across n=80 paired nights.
On-device acoustic breathing-pattern engine · by SomniAI LLC
Candidate breathing-pattern detection that runs on the device — and that you can verify yourself.
A 56.4 KB acoustic model already used inside the current SomniSense app to flag snore and breathing-pattern candidates. On-device, no cloud, with task-specific research benchmarks and an open evaluation path.
ILLUSTRATIVE stream · 03:14 PAUSE-LIKE · MODEL-ESTIMATED 22s
0:000:30now
red region = illustrative model-flagged pause-like acoustic pattern
56.4 KB
INT8 model · 9,416 params
0.064 ms
inference · on-device
n=80
person-nights · 40 participants
88.49%
breathing-window accuracy
Small is the point. The current production engine runs on-device inside SomniSense. A buyer-specific SDK boundary is extracted and validated through a scoped pilot.
Built to evaluate for
- Earbud / hearable hardware
- Mattress & bed sensing
- Smart bedside & ambient audio devices
- Remote sleep screening & telehealth
- Connected health & care platforms
Example categories for a scoped device-and-environment pilot; integration shape is defined after the operating conditions are known.
How you'd use it
The current SomniSense app uses platform-specific iOS and Android production engines to turn microphone audio into candidate acoustic fields such as labels, timestamps and confidence. A buyer pilot defines the capture contract, platform boundary and end-to-end behavior for your hardware and acoustic environment.
- On-device — runs on the phone or the chip, not a server you pay per call
- No cloud — the audio never leaves the device
- Every night — designed to run continuously, not a one-off scan
- Real-time — a live event stream, not a next-morning batch report
That last one is an opening, not a headline: a real-time stream is something your product can act on — whatever "act" means for your platform. We provide the detection layer; you own what happens next. ApneaSense is detection only; it doesn't respond, treat, or intervene.
Two ways to check us — pick the one that fits you
RIGOROUS Run it on your own data
Three technical preprints, three MIT-licensed repos. The methodology, per-seed metrics and training code are public — bring your own corpus and confusion matrix. We'd rather you trusted your own numbers than ours.
SEE IT LIVE Watch it on your own breathing
SomniSense — our live consumer app — runs the production on-device detection pipeline every night. It shows how minute-level waveforms, detected regions, audio review and whole-night timelines consume that output. The app is productization evidence; your own-device pilot remains the integration benchmark.
88.49% breathing-window accuracy snore 94.29% accuracy 56.4 KB · 0.064 ms on-device n=80 person-nights
Three proof layers — kept separate
02
Current production-engine output in SomniSense
The current app consumes candidate fields such as labels, timestamps, confidence, summaries and quality state. Those app outputs do not all inherit the window classifier's accuracy number and are not a promise of a packaged buyer SDK.
03
Your-device pilot
Your microphones, rooms, placement and users change the operating conditions. A focused pilot measures system behavior in that environment before either side treats integration performance as established.
Where it's strong, where it isn't
STRONG The reachable layer
- Cheap, on-device, real-time, zero-contact — just a microphone
- Runs every night without a clinic, a wearable, or a cloud bill
- Validated against in-lab & ambulatory PSG (n=80)
HONEST What it isn't
- An acoustic proxy — not airflow or blood-oxygen; PSG and continuous SpO₂ measure the physiological event more directly
- A screening signal, not a diagnosis; not FDA-cleared
- A precision-emphasizing threshold that can still miss or mislabel candidates; it does not guarantee a count direction for an individual night
Start with a small pilot
The default first step isn't a license — it's a focused pilot on your device, placement and environment. The pilot sets the evaluation corpus, success criteria and platform boundary before either side treats buyer integration as established.
If the agreed criteria are met, we can then discuss platform-specific SDK extraction, integration support, licensing or co-development. There is no off-the-shelf buyer package or public price tier today.
Talk to the person who built it
The email goes to the founder of SomniAI LLC — who designed and implemented the detection algorithm, is an inventor on three co-filed pending U.S. provisional patent applications, and authored the three papers and open-source repos behind it. No SDR, no sales funnel.
That means faster answers, a real technical conversation, and a direct line from your engineering lead to the person who can actually change the model.
Tell us what you're building.
Start with the papers and code, then scope a small pilot around your device and operating environment.