Applicatio ad Implemetatio of Motio Capture ad Motio Aalysis echology i Dace eachig Zhag Wei GuagXi Arts Uiversity, Naig Guagi, 5300, Chia Abstract I recet years, with the rapid developmet of computer techology, virtual reality techology as a ew product of computer techology has bee widely applied i may fields, ad motio capture ad motio aalysis as part of virtual reality are used i video, aimatio ad other scees, but the applicatio i dace teachig is rare. Dace as a kid of performace form of art, has a crucial impact o people's quality of life. his paper applies the motio capture ad motio aalysis techology i virtual reality to dace teachig, usig the three-dimesioal techology to reder realistic art, so as to let people feel persoally o the scee, thereby makig dace teachig vivid. his paper first aalyzes the status quo of Chia's dace teachig, ad the uses the classical motio capture algorithm to capture the dace actio, ad subsequetly applies the classical motio aalysis algorithm, amely feature vector method for aalysis ad matchig of captured motio. I additio, this paper also applies the model i dace teachig, greatly improvig the quality ad level of dace teachig. Key words: Motio capture; Motio aalysis; Virtual reality; Feature vector; PCA; Dace teachig 1. Itroductio Chia is both a populous coutry ad a powerful cultural coutry. Especially i recet years, with the improvemet of people's livig stadard, material life caot satisfy people's demad, ad people begi to pay attetio to the quality of spiritual life. Dace as a part of the spiritual life of people, is a kid of body movemet accompaied with musical istrumets. It is oe of the classic represetative forms of art i Chia, ad at the same time it plays a importat role i our lives. hrough dacig, studets ca fly their dream ad release the tiredess brought by learig; workers ca rela themselves ad reduce office obesity; elderly people ca achieve physical fitess ad good mood. herefore, dace for us, is omipreset. But dace learig is ot a simple thig, as a lot of dace movemets are very difficult, so the research o dace capture is immiet. his paper researches a kid of 3D model for motio capture ad motio aalysis, ad uses the motio capture i virtual reality techology to capture the dace movemets i three-dimesioal eviromet, i which the correspodig processig algorithm is used for aalysis ad processig of dace movemets, so that learers are familiar with the dace actio essetials. hus, the research has importat sigificace ad research value for the dace teachig.. Relevat research o the motio capture ad motio aalysis techology.1. Motio capture techology Motio capture is the trackig capture of the actio of the target poit by certai sciece ad techology. he motio capture techology as a emergig techology has aroused widespread cocer of researchers at home ad abroad. At preset, the research o motio capture techology i Chia still has much room for improvemet, ad this techology ot oly icludes the motio capture method, but also the real-time aalysis of movemet data. Its mai applicatio fields iclude: (1) Applicatio i three dimesioal games ad aimatio At preset, may large games ad aimatios use motio capture techology. he popular "* * game" recetly also itroduces this techology, as it eeds the real-time motio capture of teammates ad eemies to make right udgmet of attack ad defese. May famous movies also itroduce some simple motio capture techology, as show i Figure 1. 474
(a) (b) Figure 1. (a) motio capture eample ; (b) Motio capture eample () Itagible heritage protectio Motio capture is a importat way to protect the atioal itagible cultural heritage, ad our coutry has stored a series of dace movemets as data sets i the database, which ot oly protects the value of itagible cultural heritage, but also achieves sharig ad reuse of data, as show i Figure which is the diagram of the capture of some importat dace movemets. (a) (b) Figure.(a) Motio capture eample; (b) Motio capture eample.. Related research of motio aalysis techology Motio aalysis techology derives from foreig coutries, but i recet years, with the rise of emergig high techologies like data miig, machie learig, artificial itelligece ad big data, may scholars i our coutry begi to attach importace to the work of data processig ad aalyzig, while motio aalysis techology is a cocrete maifestatio of the data aalysis techology. I daily life, motio capture data i geeral are discrete ad irregular data, so it is eeded to etract the motio parameters from the motio data, while data aalysis is a importat step of the motio aalysis. Motio aalysis is a importat part of the dace teachig aalysis, ad it caot do without motio aalysis if there is oly the motio capture. Most of the 475
aalyses of actios are etractio ad aalysis based o features. here are may algorithms i this aspect. he flowchart of motio aalysis is show i Figure 3. Figure 3.Actio aalysis flow chart 3. Research o the model of motio capture ad motio aalysis algorithm his paper maily itroduces two classical algorithms of motio capture ad motio aalysis: simulated aealig particle swarm optimizatio algorithm based o PCA ad eigevector algorithm based o PCA. 3.1. Simulated aealig particle swarm optimizatio algorithm based o PCA Both the two algorithms are processig based o the PCA, but the later optimizatio process is differet though they both use PCA to achieve data pretreatmet. herefore, PCA is itroduced i detail here. Pricipal compoet aalysis is also called the PCA algorithm, which is a classical algorithm of data processig. Its priciple is to trasfer a umber of origial idicators ito a few represetative effective idicators, ad several idicators selected oly reflect most of iformatio of idicators, ad they are idepedet of each other to avoid the overlappig of iformatio, so as to achieve the purpose of dimesio reductio. his algorithm first uses PCA algorithm for dimesio reductio of motio capture data, ad the applies the simulated aealig algorithm to solve the similarity. p (1)he radom vector X is p dimesio s samples ( i i1, i,. i, i 11,,. p) he sample matri is: 476
X,,. p 11 1 1 1 1p p p 1 (1) ()he positive ide i X is ivariat, ad the adverse ide is reversed, the p (3)Make stadardizatio of elemets i Y matri ( y i y ) ui, i 1,,. p s Y. y i () y i ( y i y ) i 1 i 1 I whichy, s, by this formula, the ormalized matri is obtaied: 1 u 1 u11 u1 u1p u u 1 u u1 U (3) u u1 u up Calculate the correlatio coefficiet matri of U U U R r i (4) p p 1 I which: ( uik. uki) k 1 ri 1, i, 1,,. p (5) R I, i which, the (4)he characteristic equatio composed of correlatio matri R is 0 umber of eigevalues is P, ad the eigevalues are arraged accordig to the order from large to small (5)he cumulative variace cotributio rate is 80%~90% i geeral, ad 85% is chose as the stadard, m 1 amely 0. 85. Determie the value of m, ad the, 1,,. m. Solve the equatio p 1 Rb b, ad the the uit eigevector is obtaied: (6)Calculate the weight of the first m priciple compoet of b 0 b b i. u,., u ) ( ui1, ui zi u b 0 i, 1,, m (6) hus the pricipal compoet matri is obtaied z11 z1 z1m z 1 z zm Z z 1, z,. z m (7) z 1 z zm I which, zi is i pricipal compoets, i 1,,. m, so that the vector of P dimesio is reduced to a vector of m dimesio (7) Use the data after dimesio reductio to costruct the fitess fuctio, ad solve the similarity p ip 477
betwee the sample ad the data set. 3.. Motio aalysis algorithm of feature vectors based o PCA he feature vector s motio aalysis algorithm first searches for the attribute iformatio promotig classificatio i the origial samplig data, but this will ot chage its ow characteristic value. he algorithm is also PCA-based motio aalysis, ad therefore, there is o eed to itroduce the PCA algorithm i detail agai. For data processed through PCA algorithm, its feature vector should be obtaied. hrough aalysis of the feature vector, the correspodig aalysis of the differet movemets is achieved. Sice the feature vector algorithm is more clear ad ituitive tha the simulated aealig algorithm, ad the later is more suitable for particle aalysis, for dace actio aalysis, the selectio of PCA-based feature vector aalysis algorithm ca fully meet the requiremets of this paper. 4. Applicatio ad implemetatio of motio capture ad motio aalysis techology i dace teachig 4.1. Acquisitio of motio capture data his study uses the optical capture method commoly used i motio capture to obtai dace moves, ad actio data acquired are stored i 3D data format i the database. I the data samplig, the target wears the clothes with 1 marker poits, stadig i the specified rage, ad whe the actio capture software is opeed, the target character performs differet dace movemets, ad the camera coducts real-time trackig ad capture of the actio i the marker poit, as show i Figure 4. Figure 4. Motio capture 4.. Data storage based o Mogodb he data of marker poits collected by the camera are stored i the Mogodb database, ad because Mogodb database is a o-relatioal database, data storage format is i the form of key-value, ad it supports 3D data storage. he storage form of this database is better tha the relatioal database Mysql database, so this paper uses the Mogodb database to store the origial motio data, as show i Figure 5, which is part of the actios stored i the database. Figure 5. Database storage graph 4.3. Motio aalysis based o eigevector Actio aalysis is to get related movemet parameters of the specified dace movemets through a series of actios icludig motio trackig, capture, obtaiig ad aalysis, combie the dace movemet aalysis with dace teachig to make teachig vivid, persoalized ad iformatioized, decompose the performer s 478
movemets to achieve demostratio of these movemets oe by oe, ad make quatitative aalysis of the data obtaied. It provides a good help to the teachig of dace, ad also provides decisio support for the decisio makers. he mai steps are as follows: (1) Use the hits of optical capture to obtai skeleto iformatio of the performer i real time; () Accordig to the feature poits, 7 feature plaes are established to calculate the agular values betwee the feature vectors ad the pose vectors. Meawhile, the feature correlatio coefficiets of the movemets are calculated accordig to the movemet characteristics of the key parts of the body i the dace movemets; (3) Solve the correlatio coefficiet of the eigevector ad the icluded agle, ad aalyze the differece ad accuracy of the performer through the actio differece degree aalysis. By usig the priciple of poits movig ito lies ad lies movig ito flat plaes, three poits are selected to form a feature vector, ad the 1 markers are correspodig to 7 feature plaes. he feature vector ad agle of the mai parts of the huma body are show i table 1,. For the movemet of the limbs, the ormal vector correspodig to the limbs characteristic plae ad the ier solutio of the vertical vector are used to solve the actio s directio, ad based o the size of the icluded agle, the ormalizatio of the actio ca be udged; the head movemet ca obtai the movemet directio of the head through the compariso betwee the ormal vector correspodig to the head s characteristic plae ad the vertical vector; the movemets of the body maily iclude two movemets, amely bedig ad twistig. able 1.Discrimiat vectors o the feature plae Feature plae Left arm (P 1 ) Right arm (P ) Left leg(p 3 ) Right leg (P 4 ) Head (P 5 ) Chest (P 6 ) Belly (P 7 ) able. Discrimiat vectors o the feature plae Feature plae V 1 =V LLarm V LFarm V =V RLarm V RFarm V 3 =V Lhigh V LCrus V 4 =V Rhigh V RCrus V 5 =V LHead V RHead V 6 =V LChest V RChest V 7 =V LHip V RHip Feature plae Geometrical relatioship θ Ma θ Mi Left elbow (P 1 ) θ 1 =<V LLarm,V LFarm > 180 40 Right elbow (P ) θ =<V RLarm,V RFarm > 180 40 Left kee (P 3 ) θ 3 =<V Lhigh,V LCrus > 180 35 Right kee (P 4 ) θ 4 =<V Rhigh,V RCrus > 180 35 Head (P 5 ) θ 5 =<V LHead,V RHead > -- -- Chest (P 6 ) θ 6 =<V LChest,V RChest > -- -- Belly (P 7 ) θ 6 =<V LHip,V RHip > -- -- (4) Solve the similarity fuctio. he method of cosie similarity is preseted to solve the problem. his method ca ot oly obtai the differece betwee the vectors, but also ca get the differece ad similarity iformatio betwee icluded agles. Compared with the traditioal method of solvig Euclidea distace, this method has great advatages i the study of the differeces betwee vectors. he accuracy rate is much higher tha the error of the Euclidea distace. 4.4. Eperimetal results ad aalysis he data used i this study are from 60 studets from a uiversity i 016 that choose dace as a optioal course. he specific eperimet steps are as follows: (1) Group the studets: the studets are divided ito two groups: eperimet ad routie group through drawig lots. Because dace is the elective course, the dace level of the two groups of middle school studets is almost the same, ad there is o particularly obvious gap. herefore, the eperimetal groupig is i lie with this study. () he specific eperimetal process: the eperimetal time is from March to May, a period of two moths with 16 hours of dace teachig. For the routie group, the covetioal outdoor teachig model is applied, ad studets follow the teacher to lear dace; studets i the eperimetal group first watch 3D dace videos of actio capture, ad the practice dacig i a motio capture system, where studets ca playback ad 479
lear the specified actio accordig to their actual situatio, ad subsequetly the motio capture equipmet is used for real-time capture of studet movemet iformatio, ad the studet movemet is aalyzed by usig the eigevector method of PCA. hrough the results of the above aalysis, studets ca kow their differece with the stadard actio, so as to carry out the targeted practice. (3) Results ad aalysis. At the ed of the course, the two groups of studets dacig movemets are scored respectively, ad the average value of each group of studets is figured out. he fial eperimetal results are show i able 3. able 3. Comparative table of eperimetal results Group Movemet rage Motio itesity Motio stadardizatio Motio coherece Routie group 8 79 81 78 Eperimetal group 87 85 86 84.3 It ca be see from table 3 that the scores of the studets i the eperimetal group are sigificatly higher tha those i the routie group, whether i magitude ad itesity, or i cotiuity ad stadardizatio, so it ca be proved that the research methods used i this paper have strog referece sigificace for dace teachig. 5. Coclusio With the cotiuous developmet of techology ad social progress, sice Chia is a big coutry of educatio, there are may educators who use computer teachig methods for teachig. I this paper, motio capture ad motio aalysis techology i the computer techology is applied to the dace teachig, ad through practical eperimets, the feasibility of the algorithm is verified. hrough this method, studets costatly improve themselves accordig to their ow shortcomigs, ad dace teachers reduce their workload ad improve work efficiecy. his study promotes the developmet of dace teachig ad accelerates the research of motio capture ad motio aalysis techology i Chia o a regular basis. herefore, the research of this paper has great sigificace ad value. Refereces Araadaiel N, Villaseñor C, Lópezfraco C, et al. Structure from Motio Usig Bio-Ispired Itelligece Algorithm ad Coformal Geometric Algebra,Itelliget Automatio & Soft Computig, 017:1-7. Holz D. Power cosumptio i motio-capture systems with audio ad optical sigals,017. Kippeberg E, Verbrugghe J, Lamers I, et al. Markerless motio capture systems as traiig device i eurological rehabilitatio: a systematic review of their use, applicatio, target populatio ad efficacy:,joural of Neuroegieerig & Rehabilitatio, 017, 14(1):61. Modal R, Ray P K. Posture Aalysis of Face Drillig Operatio i Udergroud Mies i Idia: A Case Study, Advaces i Physical Ergoomics ad Huma Factors. 018. Ray S J, eizer J. Real-time costructio worker posture aalysis for ergoomics traiig,advaced Egieerig Iformatics, 01, 6():439-455. Shuai L, Li C, Guo X, et al. Motio Capture With Ellipsoidal Skeleto Usig Multiple Depth Cameras,IEEE rasactios o Visualizatio & Computer Graphics, 017, PP(99):1-1. Steves K, Huddy A. he performace i cotet model: a 1st cetury tertiary dace teachig pedagogy, Research i Dace Educatio, 016, 17():1-19. Wei X, Wa X, Huag S, et al. he Applicatio of Motio Capture ad 3D Skeleto Modelig i Virtual Fightig[C]// Iteratioal Workshop o Net Geeratio Computer Aimatio echiques. Spriger, Cham, 017:99-113. Wei Y, Ya H, Bie R, et al. Performace moitorig ad evaluatio i dace teachig with mobile sesig techology, Persoal & Ubiquitous Computig, 014, 18(8):199-1939. Xu L, Liu Y, Cheg W, et al. FlyCap: Markerless Motio Capture Usig Multiple Autoomous Flyig Cameras., IEEE rasactios o Visualizatio & Computer Graphics, 017, PP(99):1-1. Xu W, Huag M C, Amii N, et al. ecushio: A etile Pressure Sesor Array Desig ad Calibratio for Sittig Posture Aalysis, IEEE Sesors Joural, 013, 13(10):396-3934. 480