基于多特征融合的舞蹈动作识别技术研究

Dance action recognition method based on multi feature fusion

针对舞蹈动作存在复杂的姿态变化,导致舞蹈动作识别准确度低的问题。文中设计了一种多特征融合的舞蹈动作识别算法,该算法首先通过一个特征金字塔网络提取舞蹈动作的特征,再采用一个多特征融合模块来融合多路特征,以此提升算法对复杂姿态的估计,最终完成舞蹈动作识别。在数据集上的测试与分析结果表明,该算法可针对Balletto舞蹈视频数据库进行识别,且能够有效提升舞蹈动作识别准确度,从而实现对舞蹈演员的动作矫正功能。

In view of the complex posture changes in dance movements,the accuracy of dance movement recognition is low.In this paper,we design a multi feature fusion algorithm for dance action recognition.Firstly,we extract the features of dance action through a feature pyramid network,and then use a multi feature fusion module to fuse multiple features,so as to improve the algorithm's estimation of complex posture,and finally complete the dance action recognition.The test and analysis results on the dataset show that the algorithm can identify the balletto dance video database,and can effectively improve the accuracy of dance action recognition,so as to achieve the action correction function of dancers.