Related Experiment Video
Updated: Sep 27, 2025
08:05
Design and Analysis for Fall Detection System Simplification
Published on: April 6, 2020
10.8K
A Deep Convolutional Neural Network-XGB for Direction and Severity Aware Fall Detection and Activity Recognition
Abbas Shah Syed1, Daniel Sierra-Sosa2, Anup Kumar1
1Department of Computer Science and Engineering, University of Louisville, Louisville, KY 40208, USA.
Sensors (Basel, Switzerland)
|April 12, 2022
Summary
This study introduces a new system for fall detection and activity recognition using sensor data. The advanced Convolutional Neural Network (CNN) and eXtreme Gradient Boosting (XGB) model achieved 88% recall for accurate fall detection.
Area of Science:
- Biomedical Engineering
- Computer Science
Background:
- Activity and fall detection are crucial for ambient assisted living systems.
- Accurate human motion monitoring is essential for health monitoring and preventing fall-related injuries.
Purpose of the Study:
- To develop a robust fall detection and activity recognition system.
- To incorporate fall direction and severity into the detection process.
- To improve upon existing methods for recognizing daily living activities and falls.
Main Methods:
- Utilized Inertial Measurement Unit (IMU) data (accelerometer and gyroscope) from the SisFall dataset.
- Processed data into 3-second non-overlapping segments with data augmentation.
- Employed a Convolutional Neural Network (CNN) for feature extraction, followed by an eXtreme Gradient Boosting (XGB) classifier.
Main Results:
- The proposed CNN-XGB model demonstrated superior performance compared to other techniques.
- Achieved an unweighted average recall of 88% for activity recognition and fall detection.
- The system effectively classifies various activities of daily living and detects falls, considering direction and severity.
Conclusions:
- The gradient boosted CNN approach offers a highly effective solution for fall detection and activity recognition.
- This system has significant potential for enhancing safety and independence in ambient assisted living.
- Further research can explore real-world deployment and integration into smart healthcare systems.
