Accuracy of a wavelet-based fall detection approach using an accelerometer and a barometric pressure sensor.

Area of Science:

• Gerontology
• Biomedical Engineering
• Signal Processing

Background:

• Falls are a significant cause of injury and disability in older adults, with many unable to rise independently after a fall.
• Wearable sensors offer a promising solution for timely fall detection and intervention, potentially reducing the duration of 'long lies'.

Purpose of the Study:

• To evaluate the accuracy of a novel wavelet-based approach for automatic fall detection using accelerometer and barometric pressure sensor data.
• To assess the impact of sensor location and data fusion on fall detection performance.

Main Methods:

• Participants (n=15) performed simulated falls, near-falls, and activities of daily living (ADLs) while wearing sensors on various body locations.
• A wavelet transform with pattern-adapted wavelets was applied to analyze sensor data for fall detection.
• Machine learning models were used to combine wavelet and statistical features for enhanced classification.

Main Results:

• The wavelet-based method achieved high classification accuracies (82%-96%) using accelerometer data alone, with the chest sensor being most effective.
• Incorporating barometric pressure sensor data improved accuracy by an average of 3.4% (p=0.041).
• A multiphase model combining wavelet and statistical features reached a peak accuracy of 98%.

Conclusions:

• The wavelet-based approach accurately differentiates falls from non-fall events using wearable sensor data from multiple body locations.
• Combining accelerometer and barometric pressure data, along with advanced feature engineering and machine learning, enhances fall detection system accuracy.
• This technology holds potential for improving emergency response and care for older adults at risk of falling.