Research

Medication adherence is one of the leading factors that can make the difference between life and death, especially for patients managing chronic conditions. Indeed, these issues have driven a recent wave of research, including the development of smart pill bottles that monitor when a pill is extracted. In this poster, we extend our recent work (PatientSense), where we present adaptive learning techniques for subject identification across multiple pill bottle systems. We collect inertial signals from 10 subjects taking medication pills and encode the activity signals by transforming them into 2D texture images. Then we use pre-trained Convolutional Neural Network (CNN) models for image-based classification tasks. Our approach achieved improved differentiation capacity over existing models by using deep learning models, modified through domain adaptation and transfer learning.

Accurately accounting for medication use is important for the efficacy and safety of patients and family members. Monitoring is also important for medication adherence. This work investigates passive identification of persons taking medication using a sensor-equipped pill-bottle. The bottle is equipped with inertial and switch sensors in both the cap and body, making the added hardware unobtrusive, low-cost, and wireless. Our system uses inertial data to build a patient discrimination model using classification techniques. We evaluated the system using two datasets that we collected from 36 subjects. Compared to existing approaches, PatientSense is unobtrusive and can be easily used in daily life.

Smart ambient sensing applications are built on classifiers that are trained to detect different events in a physical space. Training these requires human labeling in controlled experiments. However, controlled experiments only capture specific experimental settings and application-specific labels. We aim to build a framework for data collection and active labeling that 1) reduces the number of labels necessary to maximize event coverage and 2) continuously learns the underlying distribution of different events. System Maestro is a data collection and labeling framework that senses the environment across 5 different ambient sensors producing 18 channel measurements. Maestro includes a web interface for continuous labeling and applies active learning with label propagation to minimize the number of necessary labels. We present the results of an initial deployment in a student apartment, where Maestro continuously learns to count occupants and progressively learns to identify activities of daily living.