Cognitive Errors Detection: Mining Behavioral Data Stream of People with Cognitive Impairment

Abstract

People with cognitive impairment have difficulties in planning and correctly undertaking activities of daily living due to severe deterioration in cognitive skills. As a promising solution, smart homes try to make these people live on their own with less nursing care by providing appropriate cognitive assistance while carrying out activities. For the sake of providing adequate assistance, it is necessary to understand the real intentions of residents and recognize possible anomalous trends in time during the process of performing an activity. In this paper, we analyze the abnormal behavioral patterns caused by cognitive deficits and summarize them as cognitive errors which appear frequently among people with cognitive impairment. Cognitive error detectors are designed and integrated into a unified inference engine based on Formal Concept Analysis theory. The inference engine establishes a knowledge graph hierarchically representing the interrelations between indexed activities to recognize ongoing activities, and to detect predefined cognitive errors in behavioral data streams.