Abstract
The development and construction of fall detection models represent a significant advancement in protecting health and improving the quality of life for the elderly and high-risk individuals. This study introduces a fall detection model based on images from fixed surveillance camera systems, applying deep learning models to recognize fall signs from images and videos. Several deep learning models are utilized in this research to develop fall detection technology, using image data to build intelligent recognition models. This model not only accurately and quickly identifies falls but also sends early warnings to caregivers or medical services, minimizing damage and enhancing safety. Experiments on two independent datasets, UM_Data from the University of Montréal, Canada, and LH_Data from Lac Hong University, Vietnam, show that the model achieves high performance with quick detection times and high accuracy. This research not only provides health benefits but also holds sustainable economic and social significance. Future research will focus on improving accuracy, reducing false alarms, and enhancing predictive capabilities to meet the increasing societal demand for healthcare and safety, especially for the elderly.