Related Experiment Video
Updated: Jul 2, 2026
06:49
Evaluation of a Smartphone-based Human Activity Recognition System in a Daily Living Environment
Published on: December 11, 2015
8.9K
Human Activity Recognition in a Free-Living Environment Using an Ear-Worn Motion Sensor
Lukas Boborzi1, Julian Decker1, Razieh Rezaei1
1German Center for Vertigo and Balance Disorders (DSGZ), Ludwig-Maximilians-University of Munich, 81377 Munich, Germany.
Sensors (Basel, Switzerland)
|May 11, 2024
Summary
An ear-worn motion sensor effectively recognizes human activities like walking and running with 98% accuracy using deep learning. This technology offers a practical approach for continuous health monitoring and personalized insights.
Area of Science:
- Biomedical Engineering
- Wearable Technology
- Human Activity Recognition
Background:
- Continuous behavior monitoring via Human Activity Recognition (HAR) is crucial in healthcare.
- Ear-worn sensors offer a discreet and potentially effective platform for HAR.
Purpose of the Study:
- To evaluate the feasibility of an ear-worn motion sensor for classifying daily human activities.
- To assess the performance of various machine learning algorithms, including deep learning, for this task.
Main Methods:
- Fifty healthy participants performed activities like lying, sitting, walking, and running.
- An ear-worn motion sensor collected data, which was analyzed using shallow and deep learning models (DeepConvLSTM, ConvTransformer).
Main Results:
- Deep learning models achieved 98% accuracy in classifying human activities.
- The classification models demonstrated robustness to sensor placement (either ear) and orientation variations, negating the need for calibration.
Conclusions:
- The ear is a viable anatomical location for effective human activity recognition.
- Integrating ear-worn HAR with vital sign monitoring presents a novel, comprehensive approach to personalized health assessment and tele-monitoring.
