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
Analysis of Optimal Sensor Positions for Activity Classification and Application on a Different Data Collection
Natthapon Pannurat1, Surapa Thiemjarus2, Ekawit Nantajeewarawat3
1School of Information, Computer, and Communication Technology, Sirindhorn International Institute of Technology, Thammasat University, Pathumthani 12000, Thailand. p_natthapon@yahoo.com.
Sensors (Basel, Switzerland)
|April 6, 2017
Summary
Optimal sensor placement for activity recognition is crucial. The thigh, chest, and waist positions achieved over 96% accuracy in monitoring daily living activities for young subjects.
Area of Science:
- Biomedical Engineering
- Human-Computer Interaction
- Wearable Technology
Background:
- Activity recognition systems rely on sensor data.
- Sensor placement significantly impacts model performance.
- Optimizing sensor positioning is key for accurate monitoring.
Purpose of the Study:
- To determine optimal sensor positions for activity recognition.
- To evaluate feature importance and classification algorithms.
- To provide a reference for diverse subject groups and hardware.
Main Methods:
- Investigated 19 features and 8 classification algorithms.
- Evaluated sensor positions including waist, chest, thigh, and ankle.
- Utilized Relief-F for feature selection and down-sampling for data alignment.
Main Results:
- Thigh, chest, side waist, and front waist positions yielded >96% accuracy for young subjects.
- Optimal sensor positions and feature counts were identified.
- Models validated across young and elderly datasets.
Conclusions:
- Sensor positioning is critical for accurate activity recognition.
- Specific body locations offer superior performance.
- Findings support development of robust models for diverse applications.
