摘要
针对传统舞蹈教学及训练中只能通过视觉主观评价其动作是否标准、缺少客观定量评价指标的问题,结合实际舞蹈训练应用需求,提出一种基于姿态估计的舞蹈动作相似度计算方法。该方法基于姿态估计方法实时捕捉舞蹈学习者的动作变化,并将其与标准动作进行对比,对学习者的动作是否标准做出实时判断并根据判断结果给出相应的纠正意见。首先,通过目标检测算法提取舞蹈图像中的人体检测框;然后,基于人体姿态估计算法实时估计舞蹈者的人体关节点位置信息;最后,通过设计基于三维空间关节点偏移角度变化的动作阶梯型相似性计算方法,计算学习者姿态与标准姿态之间的差异,对学习者舞蹈动作是否标准进行判断,并通过标准动作姿态还原,提出最终的动作姿态纠正意见。实验结果表明,本文方法在实时分析舞蹈者动作标准程度上具有很好的效果,在舞蹈自助教学、专业舞者动作纠正等应用场景具有一定的应用价值。
In the traditional dance teaching and training,dancers can only evaluate whether their actions are standard or not by visual subjective evaluation and there is a lack of objective quantitative evaluation index.A dance action similarity calculation method based on pose estimation considering the application requirements of actual dance training is proposed.The proposed method captures the changes of dancers′actions in real time based on the pose estimation method and compares it with the standard actions.Then,the method evaluates whether the dancers′actions are standard or not based on the dance action similarity calculation method and the corresponding correction suggestions are given according to the evaluation results.In detail,the proposed method first use the target detection algorithm to extract the human proposal in the dance image.Then the position information of dancers′body joints is estimated based on human pose estimation algorithm.Finally,a step structure dance actions similarity calculation method is designed based on the shift angle change of keypoints in the three-dimensional space and the differences between learner poses and standard poses are calculated to judge whether the learner′s dance actions are standard or not.At the same time,the final action pose correction advice is given through the standard action pose restoration.The experimental results show that the proposed method has a good effect on the real-time analysis of dancers′actions and has certain application value in the dance self-teaching,professional dancers′action correction.
作者
郭龙飞
杨红红
吴晓军
张玉梅
党允彤
GUO Longfei;YANG Honghong;WU Xiaojun;ZHANG Yumei;DANG Yuntong(School of Computer Science,Shaanxi Normal University,Xi'an 710119,Shaanxi,China;Key Laboratory of Modern Teaching Technology,Ministry of Education,Shaanxi Normal University,Xi'an 710062,Shaanxi,China;School of Journalism and Communication,Shaanxi Normal University,Xi'an 710119,Shaanxi,China;School of Music,Shaanxi Normal University,Xi'an 710119,Shaanxi,China)
出处
《陕西师范大学学报(自然科学版)》
CAS
CSCD
北大核心
2021年第5期94-100,共7页
Journal of Shaanxi Normal University:Natural Science Edition
基金
国家重点研发计划(2017YFB1402102)
国家自然科学基金(61907028)
陕西省青年科技新星项目(2021KJXX-91)。