Application of Deep Learning Models Based on EfficientDet and OpenPose in User-Oriented Motion Rehabilitation Robot Control

Abstract

This study addresses the critical challenges in rehabilitation robotics, specifically in environmental adaptability, precision in action recognition, and personalized patient care. We introduce the EfficientDet-OpenPose-DRL network, a novel integration of EfficientDet for accurate human and object detection, OpenPose for precise motion tracking, and Deep Reinforcement Learning (DRL) for optimizing rehabilitation strategies. The main contribution of this article lies in the development of this integrated model, which not only enhances environmental perception and action recognition but also improves adaptive decision-making in real-time rehabilitation scenarios. This model enables personalized, adaptable rehabilitation by leveraging advanced computer vision and deep learning techniques. Experimental results demonstrate significant improvements over existing methods, offering enhanced precision, safety, and patient-specific rehabilitation outcomes. This work contributes to the advancement of human-centric rehabilitation technologies, paving the way for more effective and interactive healthcare solutions.