Fall Detection in Smart Home Environments Using UWB Sensors and Unsupervised Change Detection

Area of Science:

• Gerontology
• Biomedical Engineering
• Signal Processing

Background:

• Falls pose significant risks to older adults' health and independence.
• Wearable sensors are common for fall detection but can be inconvenient for users.
• Alternative, non-intrusive methods are needed for reliable fall detection in home environments.

Purpose of the Study:

• To investigate the efficacy of non-wearable Ultra-Wideband (UWB) radar sensors for environmental fall detection.
• To apply unsupervised change detection methods to UWB sensor data for identifying falls.
• To evaluate the generalizability of the unsupervised approach using accelerometer data.

Main Methods:

• Utilized unsupervised change detection algorithms on UWB radar data.
• Applied the same unsupervised methods to accelerometer sensor data for comparative analysis.
• Validated the approach using real-world UWB sensor data from a Living Lab and public accelerometer datasets.

Main Results:

• The unsupervised change detection method showed promising results for fall detection using UWB radar data.
• The approach demonstrated generality by successfully applying to accelerometer sensor data.
• Non-wearable UWB radar offers a viable alternative to wearable sensors for home-based fall detection.

Conclusions:

• Non-wearable UWB radar systems present a practical and effective solution for environmental fall detection in older adults.
• Unsupervised learning methods are suitable for fall detection with UWB sensor data, offering a sensor-free approach.
• This technology has the potential to improve safety and independence for the elderly in their homes.