一个基于VR的舞蹈训练系统的研究

Research on a Dance Training System Based on Virtual Reality Environment

该研究提出了一种新型的基于虚拟现实(VR)环境的舞蹈学习系统,帮助学生可以快速识别和诊断其舞蹈动作。在该系统中,本文依据舞蹈动作所包含地独特语义进行了特征提取,并提出了基于球形自组织特征(SSOFM)神经网络的无监督的运动识别方法,将每个舞蹈动作片段的姿态分别映射到S-SOFM模型的预定义网格上的节点上,形成平滑的轨迹,依据设定的运动模版,对运动轨迹(关键姿势的序列)进行快速识别。并利用OEDTW算法计算得到输入动作和标准动作之间的形体差异度,以成绩的形式提供给舞蹈学习者。该系统框架在CAVE环境中得到实现,并通过实验验证了它的有效性与可行性。

This paper presents a new dance learning system based on virtual reality (VR) environment to help students recognize and assess their dance movements quickly. In this system,the feature extraction is carried out according to the unique semantics contained in the dance movements,and an unsupervised motion recognition method based on the spherical self-organizing feature (S-SOFM) neural network is proposed. The postures of each dance movement respectively are mapped to the nodes on the predefined grid of the S-SOFM model to form a smooth trajectory,and to identify the motion trajectory (sequence of key positions) quickly according to the set motion template. And the OE-DTW algorithm is used to calculate the difference between the input movement and the standard one,and the score is provided to the dance learner. The system framework is implemented in CAVE environment,and its validity and feasibility are verified by experiments.