Adaptive Cognitive Feedback Loop for Situation Identification in Wearable Computing Systems

Traditional Human Activity Recognition (HAR) primarily identifies basic activities, overlooking the complexity of situation identification, which depends on both activity recognition and contextual interpretation, in the light of user goals. Situationaware wearable computing systems address this challenge but often incur in high computational costs due to extensive feature sets and multiple active sensors. To optimize efficiency, we propose a feedback adaptation approach based on Endsley's Situation Awareness model, where comprehension refines perception to dynamically select the most relevant sensors and features. Experimental results on the Extrasensory dataset demonstrate improved situation identification and resource efficiency.