Berdnyk, YuriiJiayi Liu2025-07-162025-07-162025Jiayi Liu. Fall detection system for old people : Graduation project for obtaining a bachelor's degree : 123 Computer Engineering / Jiayi Liu. – Kyiv, 2025. – 173 p.https://ela.kpi.ua/handle/123456789/74972This thesis project proposes a real-time human fall detection system based on YOLO (You Only Look Once) target detection and MediaPipe posture estimation. As the trend of population aging gradually worsens, the problem of elderly falls has become an important issue of social concern. Traditional fall detection methods have problems such as response delay and inconvenience in wearing. This study combines the efficient target detection capability of YOLOv8 and the precise posture estimation technology of MediaPipe to design a monitoring and real-time feedback fall detection system. The system combines the YOLOv8 target detection algorithm and MediaPipe human posture estimation technology, and builds a friendly user interaction interface through the PyQt6 framework. The system adopts a multi-threaded architecture to achieve real-time processing and display of video streams. In the human body detection link, the system uses the YOLO model to quickly locate the human target in the picture; uses MediaPipe to extract the key points of the human body for posture analysis, and designs a fall judgment algorithm based on the characteristics of the torso angle, the height difference between the head and ankle, and the relative position of the shoulder and hip. The experimental results show that the system shows high detection accuracy and real-time performance under various environmental conditions, and the average detection accuracy can reach 90%. This study provides an effective fall detection solution for smart homes, nursing homes, medical monitoring and other fields.173 p.enfall detectionYOLOMediaPipehuman posture estimationPyQt6multithreaded processingвиявлення падіньYOLOMediaPipeоцінка постави людиниPyQt6багатопотокова обробкаBachelor Thesis