Measuring Routine Variability of Daily Activities with Image Complexity Metrics

Abstract

Since the abrupt change of daily routines can be an early symptom of cognitive impairment, it is important to measure and track the variability of daily routines of the elderly living alone in terms of their healthcare. This study is motivated by the idea that the degree of image complexity manifested in a person's day-to-day schedule chart is related to the degree of routine variability of his/her daily activities. To test this idea, synthetic data on daily activity logs containing varying degrees of routine variability was created, and a schedule bar chart image was generated based on the synthetic data. Then this study examines whether and to what extent the routine variability inherent in the dataset can be measured by existing image complexity metrics, which have been used to a pattern tendency of an image. The results indicate that the outcomes from three metrics, including Shannon Entropy, GLCM-Entropy, and GLCM-Energy, are well-associated with the degree of routine variability manifested in different aspects of daily activity schedules (i.e., start-time, duration, non-routine-contributing activities, and sequence of routine-contributing activities).