摘要
Abstract
The aim is to explore a low-cost and intelligent tennis serving action recognition and evaluation method based on Convolu-tional Neural Network,in order to provide theoretical support and practical basis for tennis technique training and analysis.Design an intelligent recognition model that combines 3D convolutional neural network,bidirectional long short-term memory network,and at-tention mechanism by constructing a tennis serve video dataset with multiple perspectives,multiple action types(flat hitting,topspin,cutting),and different technical levels,and conduct action recognition and normative scoring experiments.The proposed method achieved an accuracy of 96.8％ in the serve action classification task,which is significantly better than traditional recognition methods.Meanwhile,the model exhibits good robustness under different lighting conditions and background complexity.This method can effec-tively improve the accuracy and efficiency of tennis serving action recognition,and has great potential for application in the field of tennis technique evaluation.However,there are still certain challenges in practical competition environment application,and further optimization of model structure and generalization ability is needed.关键词
网球发球/卷积神经网络/动作识别/技术评估/时空特征建模Key words
tennis serve/convolutional neural network/action recognition/technical evaluation/spatiotemporal feature modeling分类
社会科学
