Dynamic Tracking and Real-Time Fall Detection Based on Intelligent Image Analysis with Convolutional Neural Network.

Area of Science:

• Gerontology and Artificial Intelligence
• Computer Vision and Human-Computer Interaction

Background:

• Global population aging increases demand for elder care services.
• Current caregiving manpower is insufficient to meet rising care needs.
• Real-time fall detection and analysis are critical for elder safety.

Purpose of the Study:

• To develop an AI system for real-time tracking and fall detection of care recipients.
• To enhance fall analysis accuracy using drone mobility and advanced algorithms.
• To improve the system's ability to assess various fall scenarios.

Main Methods:

• Integration of drone mobility with the Dlib HOG algorithm for real-time tracking.
• Enhancement of OpenPose for multi-person action analysis in fall scenarios.
• Development of intelligent fall posture analysis for accurate circumstance assessment.

Main Results:

• The system achieved higher identification accuracy for four fall directions compared to Google Teachable Machine's Pose Project.
• Backward fall identification accuracy improved significantly from 70.35% to 95%.
• Forward and leftward fall identification accuracy increased by nearly 14%, exceeding 95% in various scenarios.

Conclusions:

• The developed AI system demonstrates superior accuracy in detecting and analyzing falls.
• The integration of drones and enhanced AI algorithms offers a promising solution for elder care safety.
• This technology can significantly improve real-time monitoring and response to falls in aging populations.