SVM action recognition model based on skeletal key point analysis with posture sensors to help sports training

As sports and sports science evolve, tahe integration of human action recognition in sports training is becoming a crucial aspect of modern athletic development. Therefore, the study proposes an SVM-based action recognition model utilizing skeletal key point analysis with posture sensors, aiming to provide an accurate sports training analysis tool. The study employs the quaternion method to model the essential features of the human skeleton, acquires motion data through a posture sensor, and performs preliminary data processing using the Kalman filtering technique. Subsequently, it utilizes a support vector machine to complete the recognition and classification of actions. Through experimental verification, the model could effectively distinguish the feature points of different actions. The research model had a recognition accuracy of over 90% for static actions and over 80% for dynamic actions, with an average recognition accuracy of 91.24%. The results show that the human action recognition model proposed in the study has a high recognition accuracy, and its reliability and validity are verified, providing effective technical support for action improvement and technical analysis in sports training.