基于改进全卷积神经网络的体育运动员动作识别方法  

Sports Athlete Action Recognition Method Based on Improved Fully Convolutional Neural Network

作　　者：郝俊峰 HAO Junfeng(Shanxi Provincial School of Traditional Chinese Medicine,Taiyuan 030012,China)

机构地区：[1]山西省中医学校,山西太原030012

出　　处：《数字通信世界》2024年第7期55-57,共3页Digital Communication World

摘　　要：传统的体育运动员动作识别方法,直接对运动员动作识别结果进行输出未对运动区域进行提取,识别精度低。该文提出基于改进全卷积神经网络的体育运动员动作识别方法,使用摄像机对体育运动员动作图像进行采集,并对图像进行基于改进全卷积神经网络的运动区域提取,体育运动员动作识别流程,输入动作图像并对结果进行输出,实现基于改进全卷积神经网络的体育运动员动作识别。实验结果表明该研究方法识别精度高,具有一定优势。Traditional sports athlete action recognition methods directly output the results of sports athlete action recognition without extracting the movement area,resulting in low recognition accuracy.The article proposes a sports athlete action recognition method based on an improved fully convolutional neural network.By using a camera to capture images of sports athlete movements and extracting motion regions based on an improved fully convolutional neural network,the process of sports athlete movement recognition is carried out.The input action image is then outputted to achieve sports athlete movement recognition based on an improved fully convolutional neural network.The experimental results indicate that the research method has high recognition accuracy and certain advantages.

关 键 词：改进全卷积神经网络 体育运动 动作识别 识别方法 

分 类 号：G642[文化科学—高等教育学] TP393.08[文化科学—教育学]