The traditional physical education model is no longer able to meet the needs of current students, and the rise of information technology presents a new opportunity for the innovation of physical education teaching modes. This paper first introduces the application of information technology in physical education, then focuses on the information fusion technology algorithm. It constructs a multi-source information feature level fusion algorithm based on the deep typical correlation analysis algorithm and derives the multi-source feature level fusion (SVM) model from it. The authors have defined the parameters for assessing the credibility of the SVM model. Subsequently, we scrutinize the accuracy of the SVM model, and ultimately, we use this model to analyze and base our innovative opinions on the impact of information integration technology in the sports teaching model. The SVM model achieves its highest accuracy at 93.9% when R = 4, and the data follows a normal distribution with high reliability and validity, allowing for feature-level fusion. 2022 The publication rates of innovative teaching models in sports at different educational stages are 8.78%, 10.12%, 9.49%, and 8.32%, respectively, and the research of innovation in physical education teaching models is positively correlated with the year. Clearly, the innovative physical education teaching model not only utilizes big data fusion technology to tailor instruction to student’s abilities but also accurately understands their learning dynamics through remote monitoring, online learning, virtual reality, and other technologies. This serves as a model for integrating information technology into the innovation of physical education teaching models.