Spectral analysis of accelerometry signals from a directed-routine for falls-risk estimation

Area of Science:

• Gerontology
• Biomedical Engineering
• Signal Processing

Background:

• Injurious falls pose a significant risk to the aging population.
• Early identification of fall risk is crucial for preventative interventions.
• Triaxial accelerometry has shown promise in assessing fall risk through movement analysis.

Purpose of the Study:

• To improve the accuracy of fall risk assessment using accelerometry data.
• To integrate frequency-domain features with existing time-domain features.
• To validate the enhanced system in an unsupervised home environment.

Main Methods:

• Utilized triaxial accelerometry to capture movement data during a directed routine.
• Extracted both time-domain and novel frequency-domain features (harmonic ratios).
• Compared the correlation with clinical fall risk using manual and automated signal segmentation.

Main Results:

• Incorporating spectral features increased correlation with fall risk from r=0.81 to r=0.96 (manual segmentation).
• With automated segmentation, correlation improved from r=0.73 to r=0.99 after adding spectral features.
• Demonstrated a strong correlation between enhanced accelerometry analysis and clinical fall risk.

Conclusions:

• The integration of spectral analysis significantly enhances the accuracy of accelerometry-based fall risk assessment.
• The improved system shows high feasibility for unsupervised, in-home monitoring of fall risk.
• This approach supports early identification and targeted interventions for fall prevention in the elderly.