Internet of Things based pervasive sensing of psychological anxiety via wearable devices under natu ralistic settings

Psychological anxiety is highly prevalent in dementia patients, and reduces the quality of life of the afflicted and their caregivers. Still, not much technological research has been conducted to address this issue in the field. This study aimed to develop a wearable system which could detect anxiety in dementia patients under naturalistic settings, and alert caregivers via a web application. The wearable system designed included an accelerometer, pulse sensor, skin conductivity sensor, and a camera. The readings would be fed into a machine learning model to output an anxiety state prediction. One participant was trialled under both controlled and naturalistic settings. The model achieved classification accuracy of 0.95 under controlled settings and 0.82 under naturalistic settings. Implications of the study include achieving relatively high classification accuracy under naturalistic settings, and the novel discovery of movement as a potential predictor of anxiety states.