Conceptual Design of a Machine Vision-Based Corrective Feedback System for Dryland Swimming Training

Recent advances in machine vision have made real-time pose estimation possible using standard cameras. This technology holds significant potential for sports training, especially where direct coach supervision is limited. Swimming is a clear example: the aquatic environment restricts verbal communication and visual access, making dryland training a key component of technical development. However, current dryland practices lack automated corrective feedback, leaving athletes to rely on subjective self-perception.This narrative review synthesizes literature from computer vision, biomechanics, and motor learning to provide a conceptual framework for a machine vision-based system that delivers real-time corrective feedback for dryland swimming training. Three main themes are examined: pose estimation technologies (MediaPipe, OpenPose, MoveNet); feedback mechanisms grounded in OPTIMAL theory and the constrained action hypothesis; and biomechanical parameters for dryland simulation of front crawl, breaststroke, and butterfly.The findings indicate that enabling technologies are relatively mature, with upper-body tracking accuracy exceeding ۸۵% under favorable conditions and mean absolute angular errors ranging from ۳ to ۸ degrees. Theoretical foundations are also well-established. However, a major integration gap remains: no published study has systematically combined these elements into a validated system specifically for dryland swimming training. Key research gaps include the lack of validation studies for swimming-specific movements, insufficient evidence on transfer validity from land to water, and no comparative trials on optimal feedback modalities.The review concludes with a three-layer conceptual model integrating pose estimation, biomechanical rule-sets, and theory-driven feedback delivery. Future research directions include validation studies, transfer experiments, comparative trials of feedback modalities, and standardized camera protocols. This conceptual framework supports empirical work toward accessible, evidence-based technique correction for swimmers at all levels.

کلیدواژه ها:

corrective feedback ، dryland training ، machine vision ، MediaPipe ، pose estimation ، swimming technique