UrFatigue: Ultra-wideband Radar based Contactless Fatigue Monitoring

The effective monitoring of fatigue holds significant importance in maintaining long-term physiological and psychological well-being, attracting substantial attention in the field of proactive care. The existing methods for fatigue monitoring usually involve either user self-reports based on fatigue scales or the use of physiological information monitoring such as electroencephalogram (EEG) or electrocardiogram (ECG). However, these methods remain obtrusive, requiring active user participation, interfering with people's daily activities, and even adding more burden to users aiming to relieve their fatigue. In this study, we introduce UrFatigue, a passive system that monitors fatigue state via ultra-wideband radar signals, requiring neither user-initiated interaction nor the need to wear any devices. UrFatigue is built upon medically validated relationships between human fatigue and physiological/behavioral characteristics, that is, the fatigue of subjects will be reflected in physiological indicators such as heart rate and respiratory rate, as well as unconscious behaviors such as drooping head or slouching. Based on this, a radar signal feature extraction framework has been established to infer subjects fatigue. We established a prototype of UrFatigue and tested it in a simulated home environment, using artificial induction to put participants in different fatigue states. Our results demonstrate that UrFatigue has high Macro-f1 score of 91.94% in inferring a person's fatigue in a fully-automated way, providing important references for future medical applications. The source codes and radar signal dataset are public.