PhD Forum Abstract: User identification using smart pill bottles: Systems and machine learning models

Abstract

Medication adherence is one of the leading factors that can make the difference between life and death, especially for patients managing chronic conditions. These issues have driven a recent wave of research, including the development of smart pill bottles that monitor when a pill is extracted. The goal of my PhD research is to develop systems that can identify who has taken the pill and when. To do so, we have designed different generations of smart pill bottles and associated algorithms for enabling several applications. We use 3D-printed pill bottles equipped with a magnetic switch sensor and an accelerometer. The bottles are carefully designed to minimize power consumption and we devise new machine learning-based techniques that use the accelerometer data generated during bottle interaction (pill extraction) to capture the user gesture that is extracting the pill. Our work can be classified into 3 core thrust areas: 1) User identification using smart pill bottle systems. 2) Adaptive learning techniques for user identification across multiple smart pill bottles. 3) Latent conditions monitoring using smart pill bottles.