With the rapid development of science and technology, the application of wearable devices and intelligent technology in college sports soccer teaching has gradually become a research hotspot. This paper proposes a soccer action recognition and skill level assessment model based on FSR and gyroscope sensors, using FSR and gyroscope to collect pressure and angular velocity information. The discrete degree is used to determine the athletes’ leg movement state. The time domain features and frequency domain features in the collected athlete data are extracted, and the two different features are used as training samples. Ankle-based posture angle model and SVM classification algorithm model are used to recognize the movements of soccer players and evaluate their soccer skill level. The algorithm is combined with the Beidou smart bracelet. A soccer training system integrating multiple functions of localization, heart rate detection and action recognition is designed. The recognition accuracy of this paper’s method in several soccer experiments is 100%, which significantly improves the accuracy of traditional action recognition algorithms. The soccer training system provides targeted training suggestions to help students quickly master the training essentials of soccer movements, which significantly improves their soccer skill level. With further progress in technology, the system is expected to be useful in a wider range of physical education teaching scenarios.