Chapter 11 - A scalable medication intake monitoring system

Abstract

Poor medication adherence is a global issue, causing adverse health-care problems and economic consequences. Many recent studies have designed and developed medication intake monitoring systems for both directly and indirectly monitoring patients using various sensors and advanced signal processing and machine learning algorithms. However, many studies have failed to deliver a system architecture that can be easily adapted to real-life scenarios with respect to cost, size, wearability, and social acceptance. A modern system architecture for medication intake monitoring must overcome these concerns, providing a practical design that combines hardware and software to accurately identify medication intake. Furthermore, for storing and processing high-frequency sensor data streams from multiple users simultaneously, it is essential to utilize scalable data storage and computing frameworks. In this chapter, we introduce a recently developed smartwatch application and a cloud-based data pipeline. The smartwatch application collects activity sensor data and sound data using embedded inertial sensors and microphones. The cloud-based data pipeline includes distributed data storage, Apache Spark-based distributed computing, and H2O-based distributed machine learning frameworks in order to build a machine learning model that identifies instances of medication intake.