Real-Time Posture Detection and Personalized Workout Suggestion System Using MoveNet and Computer Vision

Abstract—Posture is a critical component of physical healthand overall well-being, influencing everything frommusculoskeletal integrity to athletic performance. Whileoccasional lapses in posture may not cause immediate harm,persistent poor posture can result in significant physiological andbehavioral issues over time. Addressing and correcting postureproactively is a complex challenge, making accurate assessmentand timely intervention essential for preventing long-termcomplications. The widespread adoption of connected cameras and smart devices, combined with advances in ArtificialIntelligence (AI) and Ma- chine Learning (ML), has created new opportunities for personalized, real-time posture monitoring andcorrection. Despite the clear benefits of this technologicalshift—including widespread accessibility, cost-effectiveimplementation, and proactive health management—thereremain several challenges to developing universally effectivesolutions. In this paper, we introduce a comprehensive andgeneralizable AI-based posture correction and workoutrecommendation system with the following key attributes: (i)Vision-driven: employing advanced pose estimation modelstrained on large-scale datasets for continuous posture analysis;(ii) Data-integrative: leveraging multimodal data from bodykeypoints and movement patterns; (iii) Layered: structured intoacquisition, analysis, intelligence, and feedback modules.Experimental results using real-time video data demonstrate theaccuracy and practicality of the proposed approach for postureassessment and personalized workout planning, outperformingconventional solutions.