Related Experiment Video
Updated: Aug 8, 2025
08:24
Sit-to-stand-and-walk from 120% Knee Height: A Novel Approach to Assess Dynamic Postural Control Independent of Lead-limb
Published on: August 30, 2016
10.3K
Recognizing Human Activity of Daily Living Using a Flexible Wearable for 3D Spine Pose Tracking
Mostafa Haghi1,2, Arman Ershadi1, Thomas M Deserno1
1Peter L. Reichertz Institute for Medical Informatics of TU Braunschweig and Hannover Medical School, 38106 Braunschweig, Lower Saxony, Germany.
Sensors (Basel, Switzerland)
|February 28, 2023
Summary
A flexible wearable device accurately monitors physical activity and classifies daily routines using advanced neural networks. This technology shows promise for health monitoring and rehabilitation, though fall detection requires further refinement.
Area of Science:
- Biomedical Engineering
- Wearable Technology
- Human Activity Recognition
Background:
- Physical activity significantly impacts quality of life, as recognized by the World Health Organization.
- Wearable devices offer potential for monitoring, disease assessment (e.g., Alzheimer's), rehabilitation, and fall detection.
- Non-invasive, flexible wearable sensors are crucial for unobtrusive human activity monitoring.
Purpose of the Study:
- To evaluate a flexible wearable device for 3D spine pose measurement in human activity recognition.
- To assess the device's applicability in classifying static, dynamic, and transition activities.
- To compare the performance of Long Short-Term Memory (LSTM), Convolutional Neural Network (CNN), and hybrid CNN-LSTM models for activity classification.
Main Methods:
- Developed a comprehensive activity protocol including indoor, outdoor, and transition states.
- Implemented and compared LSTM, CNN, and CNN-LSTM neural networks for activity classification.
- Utilized accelerometer and strips data for ground truth and data fusion techniques.
Main Results:
- LSTM achieved 98% overall accuracy for all activities.
- CNN with accelerometer data excelled in specific static and dynamic positions (e.g., lying down, walking, running).
- Data fusion improved accuracy for certain activities, and LSTM with strips data was effective for bending movements; fall detection accuracy reached 84%.
Conclusions:
- The flexible wearable device is effective for daily activity monitoring, recognition, and exercise supervision.
- The technology shows potential for applications in telehealth and rehabilitation.
- Further improvements are needed to enhance the accuracy of fall detection capabilities.
