Related Experiment Video
Updated: Nov 12, 2025
10:03
Conformable Wearable Electrodes: From Fabrication to Electrophysiological Assessment
Published on: July 22, 2022
4.7K
Harmonizing Wearable Biosensor Data Streams to Test Polysubstance Detection
Joshua Rumbut1,2, Hua Fang1,2, Honggang Wang3
1Population and Quantitative Health Sciences, University of Massachusetts Medical School, Worcester, MA, USA.
International Conference on Computing, Networking, and Communications : [Proceedings]. International Conference on Computing, Networking and Communications
|March 18, 2021
Summary
This study demonstrates that wearable biosensors and machine learning can detect polysubstance intoxication. Advances in wireless communication and distributed systems improve detection accuracy, with skin temperature and acceleration being key indicators.
Area of Science:
- Biomedical Engineering
- Machine Learning
- Health Informatics
Background:
- Wearable biosensors are crucial for continuous health monitoring within wireless body area networks (WBANs).
- Previous non-invasive sensor and machine learning approaches for intoxication detection have faced limitations due to inter-individual, sensor, drug, and environmental variability.
Purpose of the Study:
- To investigate how advancements in wireless communication and distributed systems can enhance polysubstance use detection.
- To develop and evaluate a machine learning model for accurate classification of intoxication from wearable biosensor data.
Main Methods:
- Harmonized two types of offline data streams from wearable biosensor readings.
- Applied supervised learning techniques to classify samples based on substance intake.
- Analyzed time and frequency domain features of biosensor data.
Main Results:
- Achieved 90% accuracy in classifying samples for polysubstance intoxication.
- Identified skin temperature and mean acceleration as the most significant predictors for intoxication detection.
- Demonstrated the potential of wireless communication advances and distributed systems in improving detection.
Conclusions:
- Wearable biosensors coupled with advanced wireless and distributed systems show promise for accurate polysubstance intoxication detection.
- Machine learning models can effectively classify intoxication states using biosensor data.
- Skin temperature and mean acceleration are critical features for developing robust intoxication detection systems.
