Year: 2025
Type: conference-paper
Abstract: In order to improve the accuracy and real-time evaluation ability of rehabilitation training, a computer vision-assisted human rehabilitation training system is constructed, which adopts end-cloud collaborative computing architecture, integrates a lightweight posture estimation model and a rehabilitation action evaluation algorithm based on time series analysis to achieve efficient data processing and feedback optimization. Experiments are conducted to analyze the effects of key point detection accuracy, movement matching accuracy and multimodal feedback mechanism on the training effect, and the results show that the system outperforms traditional methods in terms of posture estimation accuracy, real-time reasoning speed and patient compliance, and significantly improves the training standardization level in clinical trials.
Authors: Ruoyang Hu, Xiangzhou Jian, Huayu Zhao, Jun Wang
Language: English
Subfield: Computer Vision and Pattern Recognition
Field: Computer Science
Domain: Physical Sciences
Open Access status: green
OpenAlex ID: W4412624055