Conventional upper limb rehabilitation methods often encounter significant obstacles, including high costs, limited accessibility, and reduced patient adherence. Emerging technological solutions, such as telerehabilitation, virtual reality (VR), and wearable sensor-based systems, address some of these challenges but still face issues concerning supervision quality, affordability, and usability. To overcome these limitations, this study presents an innovative and cost-effective rehabilitation system based on advanced computer vision techniques and artificial intelligence (AI). Developed using Python (3.11.5), the proposed system utilizes a standard webcam in conjunction with robust pose estimation algorithms to provide real-time analysis of patient movements during guided upper limb exercises. Instructional exercise videos featuring an NAO robot facilitate patient engagement and consistency in practice. The system generates instant quantitative feedback on movement precision, repetition accuracy, and exercise phase completion. The core advantages of the proposed approach include minimal equipment requirements, affordability, ease of setup, and enhanced interactive guidance compared to traditional telerehabilitation methods. By reducing the complexity and expense associated with many VR and wearable-sensor solutions, while acknowledging that some lower-cost and haptic-enabled VR options exist, this single-webcam approach aims to broaden access to guided home rehabilitation without specialized hardware.