Detecting Infection Prevention Behaviors

Overview

During the COVID-19 pandemic, monitoring and encouraging infection preventive behaviors became critical. This project developed robust detection methods for preventive behaviors including mask-wearing, hand-washing, and disinfection using commercially available wearable devices.

Approach

• Multimodal hand hygiene detection (HCII 2023): Using audio and IMU data from off-the-shelf smartwatches to detect hand hygiene behaviors (hand washing, disinfection, face-touching) in real-world environments
• Mask-wearing detection (HealthCom 2022): Motion sensors in commercially available smartwatches to automatically detect mask-wearing status and mask types based on characteristic wearing motions
• Conversation monitoring (SmartComp 2022): "Ohanashi" system for continuously monitoring conversational events via edge processing on smartwatches

Results

• Hand hygiene detection: Achieved 93% classification accuracy for three hand hygiene behaviors using multimodal inertial-acoustic data
• Mask detection: Approximately 90% accuracy for mask-wearing motion detection and 98% accuracy for sitting/walking activity classification
• Conversation monitoring: Over 86% accuracy for conversation vs. noise classification, with 15+ hours battery life on a single charge

Significance

This research demonstrates that off-the-shelf wearable devices can serve as practical tools for monitoring infection prevention behaviors in daily life, without requiring specialized sensors or user actions.