Research On Motion Capture Data Segmentation Based On Automatic Generation Of Labanotation

Labanotation is a scientific notation,similar to the music scores,widely used to record human movements.While the motion capture data is recorded in digital form of human movements.The former is more intuitive,the image.But it’s difficult to draw.The latter is easy to get,but too abstract.Therefore,the main work of this thesis is to use the acquired motion capture data,through the computer software platform to achieve the automatic generation of Labanotation.It can be applied to the inheritance and protection of national dance,action production and so on.This thesis focuses on the motion capture data segmentation algorithm for the automatic generation of Labanotation.The main job and innovation are as follows:(1)Get motion capture data.Through the movement equipment to collect three-dimensional human motion capture data,generate BVH format files.Respectively,the part of the joint semantic and the data area are used to analysis.The semantic information of the joint is expressed by a quaternion.The azimuth information represented by the Euler angle in the data area is converted to the position coordinates.(2)The establishment of human motion standard library.For the statistical segmentation of the accuracy rate,through the establishment of human motion standard library,we can objectively evaluate the effect of data processing.At present,the standard library mainly contains the human body limbs movements,still need to follow up the perfect.(3)To improve the accuracy of motion capture data segmentation.An improved segmentation algorithm based on velocity threshold is proposed,segmentation algorithm based on probabilistic kernel principal component analysis(PPCA + KPCA),and segmentation algorithm based on feature coupling.The first segmentation algorithm is based on the speed threshold segmentation algorithm,adding filtering processing,reduce noise interference.The second segmentation algorithm is based on the idea of clustering dimensionality reduction.Combined with the PPCA(Probabilistic PCA)and KPCA(Kernel PCA)algorithm,to achieve the purpose of dimension reduction and noise reduction.The third algorithm the speed threshold segmentation processing characteristics of the curve,adding rhythm information for regular processing.By comparing with the data in the standard library,the optimal segmentation algorithm is applied to the segmentation of motion capture data.(4)The humanized attitude analysis was performed on the Labanotation data(LND)obtained by the segmentation.The establishment of the human body toward the coordinate system,according to the human joints and the coordinate axes of the offset angle,determining the location of the human body movement information,and mapping to generate the corresponding Labanotation symbols.(5)The ultimate goal of this thesis is to realize the construction of the automatic generation platform of Labanotation.The algorithm is based on the motion generation data,and the results are analyzed and compared respectively.