Achievement

We compare the results of traditional cognitive assessment with automated assessment from sensor values. Study participants include younger adults, older adults, older adults with Mild Cognitive Impairment, and older adults with dementia. Traditional assessment consists of questionnaires and lab exercises. A clinically-trained psychologist who has worked in this area for 15 years provides an assessment based on these measures. The participants then perform a set of ADL activities in our smart apartment. Using features such as time spent on the activity and distribution of sensor frequencies, our machine learning algorithms automate the assessment using lab-based outcomes as ground truth. We use an ensemble learning method where the base classifiers are trained on individual activities and the meta learner outputs the classification. The learning algorithm achieved 70% accuracy for 30 participants and a complete experiment will be performed in the coming months.