Identify sensing modalities for subtle agitation signs

Identify sensing modalities better suited for detecting facial, postural, and withdrawal signs of agitation in autistic youth than the microphones and upper-body inertial measurement units used in the study.

Background

The multimodal system’s detection performance was predominantly driven by audio and was weakest for agitation signs involving facial expression, posture, or withdrawal. These signs are poorly captured by the study’s lapel microphone and upper-body inertial measurement units. A watch-only configuration using electrodermal activity, photoplethysmography, and acceleration also performed near chance across the seven participants assessed, indicating that alternative sensing modalities may be necessary to detect less externally expressive forms of agitation.