Human fall detection on embedded platform using depth maps and wireless accelerometer.

Area of Science:

• Gerontology
• Biomedical Engineering
• Computer Science

Background:

• Falls pose a significant public health challenge for the aging population.
• Existing fall detection systems often suffer from high false alarm rates, leading to user frustration and non-acceptance.
• There is a need for cost-effective and reliable fall detection solutions.

Purpose of the Study:

• To design and implement a low-cost system for reliable fall detection.
• To minimize false alarms in fall detection systems.
• To provide a 24/7 unobtrusive fall detection solution that preserves user privacy.

Main Methods:

• Utilizing a tri-axial accelerometer to detect potential falls and user motion.
• Integrating depth map data for enhanced fall detection accuracy.
• Employing a Support Vector Machine (SVM) classifier to authenticate fall alarms after feature extraction.

Main Results:

• Development of a system capable of reliable fall detection.
• Achieved a very low false alarm ratio compared to existing systems.
• The system operates continuously (365/7/24) for unobtrusive monitoring.

Conclusions:

• The proposed system offers a reliable and low-cost solution for fall detection in aging societies.
• Combining accelerometric data with depth maps and SVM classification effectively reduces false alarms.
• The embedded system ensures unobtrusive monitoring and user privacy.