Assessing ADL Routine Variability from High-dimensional Sensing Data using Hierarchical Clustering

Abstract

Irregular patterns of Activities of Daily Living (ADLs) are associated with mild cognitive impairment (MCI) of older adults. Measuring the variability of ADL routines using various non-intrusive sensors in smart home environments presents a great opportunity for early diagnosis of MCI. However, existing studies mostly rely on supervised learning approaches to recognize ADLs and measure their variabilities, which requires large efforts in human observation and manual annotation for constructing training datasets for each home environment. In this context, this study proposes an unsupervised hierarchical clustering method to capture ADL clusters and measure their variabilities. In particular, this study focuses on addressing the challenge in employing data from multiple heterogenous sensors. The results show that the proposed method can capture the variability of ADL routines using high-dimensional non-intrusive sensing data.