Enhanced Algorithm for the Detection of Preimpact Fall for Wearable Airbags

Area of Science:

• Gerontology
• Biomedical Engineering
• Wearable Technology

Background:

• Falls are a leading cause of mortality and severe injury, particularly hip fractures, in the elderly population.
• Existing fall detection systems often lack the accuracy and timely detection needed to effectively deploy protective measures like airbags.

Purpose of the Study:

• To develop and validate a threshold-based pre-impact fall detection algorithm for wearable airbags.
• To enhance user safety by minimizing fall impact through timely airbag deployment.

Main Methods:

• Utilized inertial data, including acceleration sum vector magnitude (SVM) and angular velocity SVM, from an inertial measurement unit.
• Implemented a complementary filter with a proportional-integral controller for accurate vertical angle calculation, mitigating integration errors and external impacts.
• Recruited 30 healthy young men to simulate 6 fall types and 14 daily activities for algorithm testing.

Main Results:

• The algorithm demonstrated high performance on experimental data: 100% sensitivity, 97.54% specificity, and 98.33% accuracy.
• Achieved an average lead time of 280.25 ± 10.29 ms between fall detection and collision.
• Validated against the public SisFall dataset, yielding 96.1% sensitivity, 90.5% specificity, and 92.4% accuracy.

Conclusions:

• The developed fall detection algorithm is highly accurate and effective, even with dynamic motions.
• This technology shows significant promise for improving the safety and reducing mortality associated with falls in the elderly.
• Wearable airbags integrated with this algorithm can provide crucial pre-impact protection, preventing severe injuries like hip fractures.