Hybrid Support Vector Machine Integrated Convolution Neural Network Algorithms for Analyzing Body Motion Data for Predicting Leg Injuries

Abstract

Computer vision is also a key feature for the complete automation of applications, and various research is made to improve the performance of the prediction models to use them in better applications. Computer vision is used to summarize videos and also to analyze the activities of the subjects in the video. Accurate prediction of the subjects helps evaluate the action and provide suggestions through AI. Physical education is an essential application of computer vision, where the players' activities need to be monitored, and real-time recommendations need to be given on their practices. It helps in reducing the sports injuries of the players. This research proposes a hybrid deep learning model, which uses a CNN algorithm to predict the activities of the players and SVM to provide real-time suggestions to the players to avoid sports injuries. A model dataset is considered and used to train the model. Important physical education activities are identified, and relevant data is used to train the CNN algorithm. Based on the prediction, the SVM algorithm classifies the players' activities and provides suggestions for physical education. The model's performance is compared with earlier ones in terms of prediction accuracy. The proposed model offers 89.4 % accuracy in the prediction process for predicting different types of activities that influence physical education.