An Enhanced Ensemble Deep Neural Network Approach for Elderly Fall Detection System Based on Wearable Sensors

Area of Science:

• Gerontology
• Biomedical Engineering
• Artificial Intelligence

Background:

• Accidental falls pose significant risks of fatal injuries and hospitalizations for the elderly.
• Real-time fall detection is difficult due to the rapid nature of falls.
• Automated monitoring systems are crucial for predicting, safeguarding during, and notifying about falls in seniors.

Purpose of the Study:

• To propose a wearable monitoring framework concept for anticipating and mitigating fall-related injuries in the elderly.
• To develop and evaluate an ensemble deep neural network for real-time fall prediction and detection.
• To enhance elder care through a system that minimizes injury and improves quality of life.

Main Methods:

• Utilized an ensemble deep neural network architecture combining Convolutional Neural Networks (CNN) and Recurrent Neural Networks (RNN).
• Employed CNN for feature extraction from accelerometer and gyroscope data, and RNN for modeling temporal dynamics.
• Developed a class-based ensemble model for specific event identification (Non-Fall, Pre-Fall, Fall).

Main Results:

• Achieved high detection accuracy: 95% for Non-Fall, 96% for Pre-Fall, and 98% for Fall events on the SisFall dataset.
• The deep learning architecture demonstrated superior performance compared to existing state-of-the-art fall detection methods.
• The study validated the effectiveness of the proposed algorithm for fall prediction and detection.

Conclusions:

• The developed deep learning architecture shows significant promise for wearable fall monitoring systems.
• This approach can effectively anticipate falls, activate safety mechanisms, and issue remote notifications.
• The system has the potential to prevent injuries and enhance the quality of life for elderly individuals.