Exploring the Feasibility of Fall Detection Using Bluetooth Low Energy Channel Sounding in Residential Environments.

Area of Science:

• Engineering
• Computer Science
• Gerontology

Background:

• Falls pose significant health risks for independent older adults.
• There is a need for unobtrusive and privacy-preserving fall monitoring solutions.
• Device-free sensing offers potential for seamless integration into homes.

Purpose of the Study:

• To investigate the feasibility of using Bluetooth Low Energy (BLE) 6.0 Channel Sounding (CS) for device-free fall detection.
• To evaluate low-complexity signal representations for residential fall monitoring.
• To assess the performance of different machine learning classifiers for fall detection using BLE CS.

Main Methods:

• Two BLE nodes were used to perform periodic channel sounding, extracting scalar distance estimates.
• Time-domain and temporal-dynamic features were computed from sliding windows of the distance signal.
• Support Vector Machine, Random Forest, and XGBoost classifiers were evaluated under default and sensitivity-first operating conditions.

Main Results:

• Strong window-level sensitivity was demonstrated in subject-independent evaluations.
• XGBoost achieved the best balance between sensitivity and specificity.
• Sensitivity-first operation yielded high recall but increased false alarms.

Conclusions:

• BLE Channel Sounding is a feasible technology for privacy-preserving, device-free fall detection.
• The system offers low deployment complexity suitable for residential environments.
• Further research with larger datasets and diverse environments is needed to confirm performance.