面向拉班舞谱自动生成的运动捕捉数据分割研究

发布时间：2018-03-07 15:54

本文选题：拉班舞谱　切入点：动作标准库　出处：《北京交通大学》2017年硕士论文　论文类型：学位论文

【摘要】：拉班舞谱是一种科学的符号记谱法,类似于音乐中的乐谱,广泛应用于记录人体动作,运动捕捉数据则是以数字化的形式记录人体动作。前者比较直观、形象。但难于绘制。后者易于获取,但过于抽象。因此,本论文的主要工作就是利用获取的运动捕捉数据,通过计算机软件平台实现拉班舞谱的自动生成。将其应用于我国民族舞蹈的传承保护、动作制作等方面。本文主要研究面向拉班舞谱自动生成的运动捕捉数据分割算法,主要工作及创新点如下:(1)获取运动捕捉数据。通过运动设备采集三维人体运动捕捉数据,生成BVH文件。分别对文件中的关节语义定义部分、数据记录部分进行解析处理。采用四元组表示关节语义信息,并将数据区欧拉角表示的方位信息转换成位置坐标。(2)建立人体动作标准库。为统计分割准确率,通过手工切割建立人体动作标准库,客观评价数据处理效果。目前标准库中主要包含人体四肢动作,仍需后续完善。(3)为提高运动捕捉数据分割的准确性,提出基于速度阈值的改进分割算法、基于概率核主成分分析(PPCA+KPCA)的分割算法、基于特征耦合的分割算法。第一种分割算法是在基于速度阈值分割算法的基础上,加入滤波处理,降低噪声干扰;第二种分割算法是基于聚类降维的思想,将PPCA(Probabilistic PCA)和KPCA(Kernel PCA)算法融合处理,达到降维、降噪的目的;第三种算法对速度阈值分割处理后的特征曲线,加入节奏信息进行规整处理。通过与标准库中数据作比较,统计分割准确率,得到适用于运动捕捉数据分割的最优分割算法。(4)将分割处理得到的拉班数据(LND)进行人体姿态分析。建立人体朝向坐标系,依据人体关节点与坐标轴的偏移角度,判定人体运动部位的方位信息,并映射生成对应的拉班舞谱符号。(5)本论文的最终目标是实现拉班舞谱自动生成平台的搭建。通过编码实现基于运动捕捉数据的拉班舞谱自动生成,并对基于各种分割算法的舞谱生成效果图进行分析比较。
[Abstract]:Laban dance spectrum is a scientific notation of symbols, similar to the music score, widely used to record human actions, motion capture data is to record human movements in the form of digital, the former is more intuitive, Image. But difficult to draw. The latter is easy to get, but too abstract. Therefore, the main work of this paper is to use the captured motion capture data, The automatic generation of Laban dance spectrum is realized through the computer software platform. It is applied to the inheritance protection and motion making of Chinese folk dance. In this paper, the motion capture data segmentation algorithm for automatic generation of Laban dance spectrum is studied. The main work and innovation are as follows: 1) get motion capture data. Collect 3D human motion capture data through motion equipment, generate BVH files. In the part of data recording, the joint semantic information is represented by quaternion, and the azimuth information of Euler angle in data area is transformed into position coordinate. In order to improve the accuracy of the segmentation of motion capture data, the standard database of human body movement is established by manual cutting to objectively evaluate the effect of data processing. An improved segmentation algorithm based on velocity threshold, a segmentation algorithm based on probabilistic kernel principal component analysis (PPCA) and a feature coupling based segmentation algorithm are proposed. The second segmentation algorithm is based on the idea of clustering and dimensionality reduction, which combines the PPCA(Probabilistic algorithm and the KPCA(Kernel algorithm to reduce the dimension and noise, and the third algorithm is used to segment the characteristic curve after the speed threshold segmentation, the second algorithm is based on the idea of clustering and dimension reduction, and the algorithm combines the PPCA(Probabilistic algorithm and the KPCA(Kernel algorithm to achieve the purpose of dimension reduction and noise reduction. Add rhythm information for regular processing. By comparing with the data in the standard library, the statistical segmentation accuracy rate, An optimal segmentation algorithm suitable for motion capture data segmentation is obtained. (4) Human attitude analysis is carried out by using the Laban data (LND) obtained from the segmentation process. The human body orientation coordinate system is established, and according to the offset angle between the human body node and coordinate axis, the human body orientation coordinate system is established. To determine the position of the human body's moving position, The final goal of this paper is to build a platform for automatic generation of Laban dance spectrum, and realize the automatic generation of Laban dance spectrum based on motion capture data by coding. And based on all kinds of segmentation algorithm dance spectrum generation effect graph analysis and comparison.
【学位授予单位】：北京交通大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：J721;TP391.41

【参考文献】