THE ANALYSIS OF CHILDREN’S SPORTS BEHAVIOR ANALYSIS AND HEALTH MANAGEMENT SYSTEM BASED ON DEEP LEARNING

In recent years, with the change of lifestyle, the problem of children’s obesity and physical decline has become increasingly serious, which poses a threat to the healthy growth of children. Therefore, the scientific management and effective analysis of children’s sports behavior is particularly important. The traditional analysis method of children’s sports behavior mainly relies on manual observation and recording, which is not only inefficient, but also easily influenced by subjective factors, and it is difficult to ensure the accuracy and objectivity of the analysis. In view of this, this study aims to develop a children’s sports behavior analysis and health management system by using deep learning technology to improve the accuracy and efficiency of sports behavior analysis. The significance of this study is that it cannot only provide parents and doctors with a scientific health management basis, but also provide data support for formulating more scientific children’s health policies by deeply understanding children’s sports habits and needs. The purpose of this study is to build a children’s sports behavior analysis and health management system based on deep learning (DL), which can automatically identify and analyze children’s sports behavior and provide personalized health guidance. The system uses You Only Look Once version 7 (YOLOv7) DL model to automatically identify and analyze the movement behavior, and improves the performance of the model by integrating the attention mechanism and optimizing the loss function. The research method is to automatically identify and analyze children’s sports behavior by collecting children’s daily sports data and using the DL model YOLOv7. In order to ensure the generalization ability of the model, a group of children with different ages, sexes, and body shapes were selected as the experimental objects. YOLOv7 model is trained by using pre-trained weights from large public data sets, and the Adam optimizer is used to minimize the loss function. In order to improve the training efficiency and model performance, grid search and cross-validation are used to adjust the superparameter. The main conclusion of the study is that the designed YOLOv7 model has achieved high accuracy, recall, and F1 score in children’s motor behavior recognition tasks. Specifically, the accuracy of the model in running category is 95.0%, the recall rate is 93.0%, and the F1 score is 94.0%. The accuracy rate of playing basketball is 89.0%, the recall rate is 86.0%, and the F1 score is 87.5%. These results show that the designed YOLOv7 model has excellent performance in the task of children’s motor behavior recognition, which proves the effectiveness and accuracy of the system. In particular, by introducing innovative designs such as attention mechanism and optimizing loss function, the performance of YOLOv7 model is further improved, so that it can better adapt to the task of children’s sports behavior recognition.