This is a submission for the Presage SmartSpectra SDK Challenge
What I Built
Children on the autism spectrum often experience distinct changes in their breathing and pulse due to differences in how their autonomic nervous system (ANS) regulates automatic bodily functions.
Because their brains process sensory input (like loud noises, bright lights, or textures) and emotional shifts more intensely, their bodies hit the "panic" button faster and have a harder time hitting the "calm down" button.
Many autistic children experience differences in interoception—the ability to sense what is happening inside their bodies. They may not realize their heart is racing or their breathing is shallow until they are already on the verge of a meltdown or shutdown.
Building an early-warning alert app with the Presage SmartSpectra SDK is a practical approach because contactless camera sensing eliminates the sensory discomfort of chest straps or smartwatches for children on the spectrum.
CalmSense is a contactless monitoring and early-warning app built with the Presage SmartSpectra SDK to help parents and caregivers spot the early signs of sensory overload or an approaching meltdown in children on the autism spectrum—without requiring uncomfortable wearables or chest straps.
The Idea - How It Works
Contactless Camera Sensing: Using a standard camera or video feed, CalmSense tracks subtle autonomic nervous system shifts—including pulse rate, heart rate variability (HRV/RMSSD), Baevsky Stress Index, and shallow vs. deep breathing patterns—alongside early behavioral cues like rocking, pacing, or moving hands toward the ears.
Personalized 3-Stage Detection: Instead of relying on rigid medical averages, the app compares live metrics against each child's calibrated calm baseline across a 3-stage escalation model (Early ANS Arousal → Sensory Escalation → Acute Overload) to minimize false alarms.
Proactive Parent Alerts & Auto-Remediations: When sustained escalation is detected, CalmSense sends an early alert to the parent with suggested calming actions and can automatically trigger soothing smart-home routines—such as dimming lights, playing calming music, launching a favorite comfort show on TV, or starting an Alexa quiet-time routine—before a full episode occurs.
Demo
Code
The video ( test_video.mp4) used in the code is not uploaded, go to the youtube link and download the video and put in the the CalmSense folder when you clone the repo (also put it in web folder, if you want to see the video on the web page)
Public Code Repo
viveknambiar / CalmSense
**CalmSense** is a contactless monitoring and early-warning prototype built with the **Presage SmartSpectra C++ SDK** and **OpenCV**
CalmSense: Contactless Pre-Episode Detection & Early-Warning System
CalmSense is a contactless monitoring and early-warning prototype built with the Presage SmartSpectra C++ SDK and OpenCV. It helps parents and caregivers detect early physiological and behavioral signs of sensory overload or an approaching meltdown in children on the autism spectrum—without requiring uncomfortable wearables or chest straps.
Key Features
- Contactless Camera Sensing: Extracts pulse rate, breathing rate, Heart Rate Variability (HRV/RMSSD), Baevsky Stress Index, and blink cues from a standard video feed.
- Hand-to-Ear / Facial Occlusion Detection: Uses OpenCV YCrCb skin-masking and frame-differencing across upper ear/temple regions to detect when hands are lifted to cover the ears.
- Personalized 3-Stage Pre-Episode Signature:
- Stage 1 (Early ANS Arousal): HRV (RMSSD) drop > 15%, rising Stress Index, or blink-rate spikes.
- Stage 2 (Sensory Escalation): Shallow/rapid breathing paired with rocking/pacing cues.
- Stage 3 (Acute Overload): Simultaneous pulse and breathing spikes or hands lifted to cover ears.
- …
How It Responds to Human Signals
Prize Categories
Best Thoughtful Real-World Use
To predict an approaching meltdown, shutdown, or sensory overload episode in a child on the spectrum, you need to look beyond raw Pulse Rate and Breathing Rate (which spike late, often when the episode is already underway).
Instead, combine SmartSpectra’s deeper autonomic nervous system (ANS) biomarkers with early behavioral micro-cues visible in the camera frame.
Right now the demo prototype is a Web App, but this can be developed as a phone app.
My favorite feature is Auto Remediation. By connecting the app to your smart home devices—like dimming the lights, playing music through Alexa, or turning the TV to a favorite show—it can automatically adjust the environment to help calm the person down.
Consent and Data Handling
I confirm that the video used in this demo is my own original recording, filmed at home on my personal desktop. I am the sole individual appearing in the footage, which is hosted on a public YouTube channel, and I explicitly consent to its use for running this demonstration code.
