Research On Automatic Generation Of Labanotation Based On Dynamic Programming

Labanotation is a kind of action notation as similar as staff notation of music.Moreover,it is intuitive,convenient and easy to read.So Labanotation has been widely recognized and used.It can record any simple or difficult motions scientifically and has become an internationally recognized dance communication tool.Due to careful analysis,drawing of a Labanotation is time-consuming and laborious.Motion Capture is a concept originated from the animation technology of twentieth Century.Modern Motion Capture technology has been quite mature,and has been widely used in film and animation.Motion capture make animation very realistic.And capture data is more and more accurate day by day but has high equipment costs.This paper aims to generate Labanotation automatically with motion recognition technology of computer.We will take motion capture equipment as eyes and then analyze the data to obtain the action recognition results.So as to improve the record efficiency of Labanotation.The research provide another way for the protection of national folk dynamic art.This paper introduces a method of automatic generation of Labanotion based on dynamic programming.The format of the capture data is BVH(Bio-vision,Hierarchical).Through the understanding and analysis of the data,we do basic motion segmentation and recognition.And at last we finished the Labanotation generation platform.First,a passive optical motion capture system is used to collect human motion data,saved as BVH file.Analysis of the definition of the bone hierarchy have been done.And match the node and the semantic of human body joints.The data in the BVH format is then converted into easy-to-use position coordinate data for subsequent analysis.Second,a method based on kinematics feature is used to segment continuous motions into basic motions.In order to improve the accuracy and take advantage of dance motions,we add some rhythm information to do a time warping on the segmentation results.Before motion recognition,another import work have been done.A template library with 110 element motions and a sample library with 1107 element motions were built.Then two method were applied to the motion matching.The one is based on key frame sequence similarity matching while another one is based on the Dynamic Time Wraping.And the accuracy is tested on the sample library.At last,the method based on Dynamic Time Wraping with higher accuracy rate is applied to the platform.In order to output the Labanotation,we introduce a special data structure to present the Labanotation symbols,thus we can convert the data into Labanotation.In summary,a system for automatic generation of Labanotation from Motion Capture Data is proposed in this research.Furthermore,the platform of this system has been developed with python.The Labanotation generated by the platform compared with the Labanotation recorded by the specialist is correct.So the method achieve the goal of automatic generation of Labanotation.