AMMF-RC: An Adaptive Multi-Modal Model for Real-Time Fitness Motion Recognition and Personalized Correction

As the demand for intelligent technology in the fitness field continues to grow, traditional fitness guidance systems often lack sufficient accuracy and timeliness in action recognition and real-time feedback. To address the challenges of multi-modal data fusion and personalized real-time feedback, this paper proposes an adaptive multi-modal real-time fitness motion recognition and correction model (AMMF-RC). Integrating visual and sensor data, the model enhances action recognition accuracy through multi-modal data fusion and provides personalized correction recommendations based on an adaptive feedback mechanism linked to user status. Experimental results demonstrate that AMMF-RC achieves significantly higher recognition accuracy across various fitness actions compared to single-modality and traditional multi-modal models, with a considerable reduction in feedback latency, effectively enhancing user training experiences. AMMF-RC not only expands the application scope of fitness guidance systems but also offers a viable reference framework for future intelligent fitness system development.