Rehabilitation Exercise Quality Assessment through Supervised Contrastive Learning with Hard and Soft Negatives

arXiv:2403.02772v1 Announce Type: new
Abstract: Exercise-based rehabilitation programs have proven to be effective in enhancing the quality of life and reducing mortality and rehospitalization rates. AI-driven virtual rehabilitation, which allows patients to independently complete exercises at home, utilizes AI algorithms to analyze exercise data, providing feedback to patients and updating clinicians on their progress. These programs commonly prescribe a variety of exercise types, leading to a distinct challenge in rehabilitation exercise assessment datasets: while abundant in overall training samples, these …

abstract ai algorithms algorithms analyze arxiv assessment cs.ai cs.cv cs.cy cs.lg data exercise feedback home life mortality patients quality through type virtual