딥러닝 기반의 사용자 맞춤형 홈트레이닝 로봇 시스템
Deep Learning-Based Personalized Home Training Robot System
0
조회
0
다운로드
0
인용
0
참고문헌
키워드
초록
This study presents a novel system for real-time exercise pose recognition, repetition counting, and user position optimization using deep learning-based keypoint extraction and autonomous robot control. The system captures live video data from a robot-mounted camera and leverages the YOLOv8 model to extract key body landmarks, such as the head, shoulders, elbows, knees, and wrists. Based on these landmarks, the system calculates joint angles to determine three exercise pose types: squats, lunges, and sidearm. The robot autonomously adjusts its position to maintain an optimal distance and alignment with the user, ensuring precise pose detection. When the user maintains a specific angle for a set duration, the pose is confirmed, and repetition counting begins. Each correct movement increments the count, while incorrect postures trigger real-time voice feedback to guide the user. Upon completing a predefined number of repetitions, the system signals workout completion and prepares for the next exercise. This research enhances personalized exercise support systems by integrating real-time pose analysis, autonomous user tracking, and feedback mechanisms for improved accuracy and usability.
Abstract
This study presents a novel system for real-time exercise pose recognition, repetition counting, and user position optimization using deep learning-based keypoint extraction and autonomous robot control. The system captures live video data from a robot-mounted camera and leverages the YOLOv8 model to extract key body landmarks, such as the head, shoulders, elbows, knees, and wrists. Based on these landmarks, the system calculates joint angles to determine three exercise pose types: squats, lunges, and sidearm. The robot autonomously adjusts its position to maintain an optimal distance and alignment with the user, ensuring precise pose detection. When the user maintains a specific angle for a set duration, the pose is confirmed, and repetition counting begins. Each correct movement increments the count, while incorrect postures trigger real-time voice feedback to guide the user. Upon completing a predefined number of repetitions, the system signals workout completion and prepares for the next exercise. This research enhances personalized exercise support systems by integrating real-time pose analysis, autonomous user tracking, and feedback mechanisms for improved accuracy and usability.