Discriminating and Measuring Activities of Daily Living

Activities of Daily Living (ADLs) are used to assess the ability of individuals to look after themselves in the home, and can be used to aid the diagnosis of degenerative brain diseases, such as Alzheimer's Disease. The current methods for assessing ADLs require the subjective input of a human, and as such may not reflect the objective capabilities of the individual. Therefore, this work explores the automatic and objective quantification of ADLs using data obtained from a wearable medical device which records eye- and head-movements. We focus on detecting four basic activities including eating, brushing teeth, walking and inactivity. Using a subject-independent testing framework, in which the data from each individual being tested does not appear in the training data, our main contribution is to explore a range of contemporary approaches to time series classification and to compare them to our own previous work in this area. Using a MUSE classifier, we obtain a peak mean accuracy of 84.72%. We also present a system for classifying the duration of each activity, which we suggest may be a good parameter for determining how well an activity has been performed. In this task, we achieve perfect discrimination between three durations of 30 s, 60 s and 90 s.Clinical relevanceDetailed measurements of activities of daily living could aid the diagnosis, monitoring and care of individuals with degenerative brain conditions.