ENHANCING PHYSICAL EDUCATION MOVEMENT PRECISION THROUGH AI-DRIVEN DEEP LEARNING CALIBRATION SYSTEMS

Abstract

The aim of this study is to enhance the efficacy of sports teaching movements and to promptly correct erroneous forms, by integrating artificial intelligence (AI) and deep learning technologies into the recognition of sports movements. This paper commences by computing the correlation matrix for a set of selected features, subsequently establishing a threshold to eliminate features with high cross-correlation, thereby reducing redundancy and optimizing the feature set. To preprocess the imagery, a Gaussian function is initially applied to perform convolution operations. Subsequently, a Gaussian kernel function is utilized to filter the images, constructing a hierarchical structure known as the Gaussian pyramid, wherein variable Gaussian filter coefficients are employed at each level of image processing. Ultimately, this research develops a precise calibration system for physical education movements and implements it within the context of physical education to enhance teaching outcomes. The experimental results demonstrate that the system developed in this study effectively satisfies the practical requirements of physical education.