基于多尺度特征融合的舞蹈规范动作姿态估计方法  

Motion pose estimation method of dance norm based on multi-scale feature fusion

作　　者：常丽萍 CHANG Li-ping(Wenzhou University,Wenzhou,Zhejiang 325035,China;Fuyang Preschool Teachers College,Fuyang,Anhui 236015,China)

机构地区：[1]温州大学,浙江温州325035 [2]阜阳幼儿师范高等专科学校,安徽阜阳236015

出　　处：《河北北方学院学报（自然科学版）》2022年第11期15-22,共8页Journal of Hebei North University:Natural Science Edition

基　　金：2021年度安徽省教育科学研究项目:“‘双增’‘双减’政策下少儿舞蹈融入小学课后延时服务的价值研究——以阜幼专附属小学为例”(ZLGC2021JY003)。

摘　　要：舞蹈表演时的肢体动作灵活、表现方式复杂,姿态估计准确率较低。为此,提出了一种基于多尺度特征融合的舞蹈规范动作姿态估计方法。设计PCA分割模型,当增加至某数据帧时,提取出的主元成分与前一个区间段中的成分存在较大差异,由此确定此时的动作已经发生改变,实现动作序列分割。基于HSV空间设计一种视频前景分割算法,分割序列分割结果的视频前景。结合深度可分离卷积网络与级联金字塔网络构建舞蹈规范动作姿态估计模型,并生成舞蹈规范动作图像,实现舞蹈规范动作姿态估计。测试结果表明,该方法的姿态估计准确率在独舞时高于95%,群舞时高于92%;在明亮光线下高于93%,阴暗光线条件下高于86%,整体姿态估计准确率较高,满足了最初设计目标。Due to the flexibility and complexity of body movements in dance performance,the accuracy of pose estimation is low.Therefore,a dance standard pose estimation method based on multi-scale feature fusion is proposed.The PCA segmentation model is designed.When the data frame is added to a certain data frame,the extracted principal component is greatly different from the component in the previous interval segment,so it is determined that the action at this time has changed,and the action sequence segmentation is realized.A video foreground segmentation algorithm based on HSV space is designed to segment the video foreground of sequence segmentation results.Combined with the deep separable convolutional network and the cascade pyramid network,the estimation model of dance standard movement pose was constructed,and the dance standard movement image was generated to realize the estimation of dance standard movement pose.The test results show that the accuracy of pose estimation is higher than 95%in solo dance and 92%in group dance.It is higher than 93%in bright light,higher than 86%in dark light,and the overall attitude estimation accuracy is high,which meets the initial design goal.

关 键 词：多尺度特征融合 PCA方法 视频前景 舞蹈规范动作 姿态估计 

分 类 号：TP311.13[自动化与计算机技术—计算机软件与理论]