This research shows how a wearable can learn when an autistic child’s body is shifting into an anxious state—just from their heart rhythm, in real time. It’s built for kids who may not be able to say “I’m anxious,” but whose physiology clearly changes when they are.

Instead of needing hours of labelled “calm vs anxious” training data for every child, the algorithm continuously learns that child’s personal “calm” heart‑rate pattern and then looks for meaningful deviations—like a sustained, anxiety‑type increase in heart rate. It does this using a mathematical tool called a Kalman filter, which predicts what the next heartbeat interval should be and flags when reality consistently looks more “stressed” than expected.

Tested on autistic children doing anxiety‑provoking tasks (like a Stroop test and public speaking), the system detected anxiety‑related arousal with about 99% sensitivity and 92% specificity—very few misses and very few false alarms. In plain language: this is a fast, accurate, training‑light way for wearables to spot when anxiety is rising, so your product can step in with support at the right moment