Application of Adaptive Virtual Reality with AI-Enabled ...

Research ArticleApplication of Adaptive Virtual Reality with AI-EnabledTechniques in Modern Sports Training

Yajun Zhang 1 and Sang-Bing Tsai 2

1Department of Physical Education Henan University of Animal Husbandry and Economy Henan 450046 China2Regional Green Economy Development Research Center School of Business Wuyi University Nanping China

Correspondence should be addressed to Yajun Zhang pandajunjun03126com and Sang-Bing Tsai sangbinghotmailcom

Received 9 August 2021 Revised 31 August 2021 Accepted 7 September 2021 Published 18 September 2021

Academic Editor Han Wang

Copyright copy 2021 Yajun Zhang and Sang-Bing Tsai -is is an open access article distributed under the Creative CommonsAttribution License which permits unrestricted use distribution and reproduction in anymedium provided the original work isproperly cited

Virtual reality technology has many characteristics such as high immersion dynamic interactive response and multidimensionalinformation digitization -ese characteristics are needed by the combination of modern sports development and scientific andtechnological progress -e introduction of virtual reality technology into the field of sports will be conducive to the scientifictraining of sports competition and the digital research of technical theory and to the rapid development of modern sports -ispaper constructs the framework of sports training based on virtual reality technology and designs a motion capture data algorithmbased on behavior string which successfully improves the advantages of virtual reality technology In addition the trainingexperiment method is used to verify the effectiveness and superiority of the system focusing on the high-school students learningtennis for the first time as the experimental object to study the differences in training effects between the traditional trainingmethod and the training method using virtual reality technology-e training effect of tennis is better than the traditional trainingmethod and the difference is significant the internal motivation of training psychology is better than the traditional trainingmethod and the difference is significant Compared with the traditional training method this training method can stimulatestudentsrsquo interest in training improve studentsrsquo training effect and promote studentsrsquo psychological internal motivation tocontinue training -is study provides useful enlightenment for the further application of virtual reality technology and in themodern sports training for various sports events

1 Introduction

Since the twentieth century human society has entered theinformation age Modern science and technology has a trendof high differentiation and high concentration and has alsoentered a new stage of multidisciplinary comprehensive uti-lization [1 2] A large number of modern research methodsand technologies have been transplanted to sports trainingand scientific research Many achievements of modern hightechnology have also begun to penetrate into the field ofcompetitive sports and have had a series of profound effectson modern competitive sports training [3] In particular therapid development of virtual reality technology provides abroader space for the application of high technology in thefield of competitive sports [4] Summarizing the successful

experience of Beijing Olympic Games it is found that usingvirtual reality technology for simulation training can realizethe human motion measurement method from traditionalhuman eye observation to high-precision motion capture andanalysis [5] -e two changes from the experience-basedmethod to the human motion analysis method of humanmotion simulation and simulation will be faster and moreeffectively improve Chinarsquos training level and competitivelevel in these projects and realize the magic weapon of theOlympic glory plan In order to ensure Chinarsquos dominantposition in the total number of gold medals and medals in theOlympic SummerOlympic Games it is particularly importantto timely track the development of virtual reality technologyand deeply study the practical significance of virtual realitytechnology to the development of competitive sports [6]

HindawiMobile Information SystemsVolume 2021 Article ID 6067678 10 pageshttpsdoiorg10115520216067678

Some researchers are committed to the application ofvirtual reality in physical training [7] Akbas et al [8] holdthat ldquothe application of virtual reality technology in physicaleducation includes five aspects knowledge learning virtualsports experiment skill training physical education networkdistance teaching and academic exchange Wei et al [9] andothers proposed that the use of virtual reality technology canmake physical education courseware and carry out physicaltraining Lohre et al [10] hold that the application of virtualreality technology in physical education includes five as-pects physical education teaching and training makingphysical education network distance teaching accidentprevention and academic exchange Arrighi et al [11] holdthat ldquothe application of virtual reality technology in collegeteaching includes virtual campus virtual classroom virtualexperiment and FGH courseware designWu and Zhou [12]put forward that ldquothere are four points in the application ofvirtual reality technology in Physical Education Teachingone is to simulate sports scenes and break through thelimitation of time and space second innovate the educa-tional model to improve studentsrsquo interest in learning andteaching effect third digital learning tracking learning dataand effectively monitoring learning behavior fourth it hasstrong functional development and broad applicationprospects in college physical education [13]rdquo Synovec et al[14] hold that the application of computer-based ldquovirtualrealityrdquo technology in college sports training has the fol-lowing four points first the comparison between sportstechnical action and reality is completed through ldquovirtualrealityrdquo technology the second is to use ldquovirtual realityrdquotechnology to realize the comparison between virtual actionand virtual reality third build different training environ-ments through ldquovirtual realityrdquo technology fourth the use ofcomputer ldquovirtual realityrdquo technology can realize remoteinteractive training [15]

-e application of virtual reality technology in collegephysical education teaching and training is the main trend ofcollege physical education development It can effectivelystimulate studentsrsquo sensory cognition and thinking modeand plays an important role in strengthening the effect ofstudentsrsquo physical education learning and training -ere-fore the contributions of this paper are as follows (1) designthe motion training framework based on virtual realitytechnology (2) design the motion data capture algorithmbased on behavior string to capture the motion trajectory ofthe target (3) the control group was designed and comparedwith the traditional training methods to verify the advan-tages of the modern training system based on VR -roughthe research and understanding of virtual reality technologyfind out the advantages of virtual reality technology andmake it flexibly applied to high-school tennis training whichcannot only enrich the training content and training meansbut also improve the training efficiency and training effect oftennis class and create a good learning atmosphere andlearning environment so as to improve the training quality-erefore this paper constructs the framework of sportstraining based on virtual reality technology and designs amotion capture data algorithm based on behavior stringwhich successfully improves the advantages of virtual reality

technology With the improvement of studentsrsquo interest inlearning the mastery of tennis skills is also improving whichis conducive to the further research of tennis training theory

2 Structure and Design of Sport TrainingBased on Virtual Reality

21 Architecture Design of Cloud Platform According to theproblems existing in the current sports training system basedon virtual reality the motion capture technology is intro-duced into the human motion posture simulation andanalysis system [16]-e data of the files obtained by motioncapture are analysed by the least square method and theresults are applied to the simulation system which will helpto solve many problems and deficiencies in the humanmotion posture simulation system [17 18] For the researchwork of using motion capture technology to obtain data andextract human motion posture parameters the developmentof this system is completed with the support of Bessel curvetheory -e physical training system based on virtual realityis mainly composed of three functional modules data ac-quisition module motion posture capture module interfacemanagementmodule and system kernel module-e overallframework of the sports training system based on virtualreality is shown in Figure 1

-e physical training system based on virtual realityadopts a hierarchical architecture which is divided into uservisualization layer signal processing layer and signal ac-quisition layer from top to bottom -e user visualizationlayer is responsible for providing a ldquocontainerrdquo for allfunctional modules receiving various messages provided byusers and transmitting the received information to the lowerlayer that is the signal processing layer so as to completethe interaction between users -e user interface module is aportable module of the presentation layer which is mainlyused to present the user interface with certain personalityand style Although the user interface module is relativelysingle and the personality is not very distinctive it has verygood interaction and powerful function and can completethe basic functions required by the system -e threemodules of model Coleman filter and human motionposture simulation in the signal processing layer belong tothe system core -e data collected by 3D suit motioncapture equipment are applied to human motion posturesimulation and combined with Bezier curve and numericaldata to complete realistic humanmotion posture simulationthe realistic simulation of the whole human motion postureis completed in these three modules the Bezier curvemodule is an extension module of the logic layer It is re-sponsible for freely editing curves to achieve the purpose ofintuitive and realistic simulation of human motion -emotion data acquisition module is combined with themotion data acquisition module of the human body modelto achieve the purpose of reading the motion data of thehuman body model

Practice has proved that with the hierarchical archi-tecture the system has clearer hierarchy lower coupling andstronger scalability -erefore the system is easy to main-tain improve and expand -e independence of modules

2 Mobile Information Systems

makes it easy to modify replace or reuse modules in eachlayer As long as the interface between modules remainsunchanged a layer can also provide a variety of inter-changeable specific implementations -e design supportedby the hierarchical system reflects the increasing level ofabstraction simplifies complex problems into a series ofsimple incremental steps and makes the system highlyscalable and flexible

22 Overall Design of Whole Body Motion Capture SystemIn order to improve the accuracy of human motion capturedata the system integrates the data of two Kinect devices[18 19] and adopts the client-server mode and each device isconnected with the corresponding client PC Two clientcomputers use SDK to obtain skeleton data and send it toserver-side PC through local Ethernet All bone data areprocessed in the server and then the fused bone data istransmitted to the Unity3d scene for visualization [20]Because Kinect has the disadvantage of self-occlusion itcannot obtain the full picture of human posture In thispaper two Kinect devices are taken as an example to form adual Kinect human motion capture system which canobtain more accurate human motion data through datafusion and expand the range of human activities Whenplacing Kinect try to obtain the whole picture of the humanbody -e total viewing angle of the two Kinect to obtaindata is as large as possible so the orthogonal layout scheme isadopted At this time the mutual interference of infraredrays is also the smallest According to the height of thehuman body and the best capture range of Kinect the twoKinect are placed at a vertical height of 70 cm from theground and 2m away from the fusion centre -e systemscheme design is shown in Figure 2

In addition in order to realize the data transmissionbetween the client and server it needs to be completedthrough data transmission protocol [21] -e commonlyused communication transmission protocols of the twohardware devices include TCP protocol and UDP protocol-e client converts the bone and joint data obtained fromKinect sensor into OSC information format for transmissionand sends it through UDP protocol -e server takes Uni-ty3d engine as the main body to realize data fusion andvisualization and adds 0sc component to Unity to receive thedata sent from the client Set the IP address and the ports forsending and receiving information establish a connectionbetween the client and the server connect the client and theserver PC to the same LAN at the same time and completethe data transmission through the UDP protocol

3 Motion Data Capture Research Based onBehavior String

31 -e Process Design for Behavior String Capture -epurpose of motion data behavior segmentation is to auto-matically separate human motion capture data containingdifferent motion types to form motion fragments with in-dependent semantic information which is convenient forstorage and reuse in the database and lay a foundation formotion analysis and synthesis Amotion training frameworkbased on virtual reality technology is constructed and amotion capture data algorithm based on behavior string isdesigned -e motion capture data algorithm based onbehavior string can encode and reconstruct the motiontrajectory according to the previous motion frames to form abehavior string and then capture the target motion dataindirectly according to the historical behavior string whichcan significantly improve the accuracy of motion capture

Head motion

MCU

Voice recognizer

Bessel curve

Computer

Signal acquisition module Application Visualization

Motion attitude simulation

Kalmanfiltering

Motion capture

User visualization

Signal processing module

Figure 1 Structure design for sport training based on virtual reality

Mobile Information Systems 3

-is paper proposes a new string representation methodbased on human motion capture data which represents thehigh-dimensional human motion capture data sequence inthe form of string In this method the human motioncapture data is regarded as a high-dimensional data pointset the density-based clustering method w is used forclustering and the obtained categories are represented bycharacters -rough time sequence recovery processing theoff-dimensional data point set represented by characters isreordered according to the time sequence of the originalmotion sequence to obtain the string corresponding to theoriginal motion capture sequence and this string is calledbehavior string (BS) Finally by analysing the behaviorstring the behavior segmentation of the human motioncapture data sequence is realized and the motion cyclecorresponding to various behaviors is extracted

In this paper the distance between the human bodymodel and the local motion points of the human bodymodelis regarded as the similarity between the two motion pointsof the human body model and the distance between the twomotion points and the virtual human body model is cal-culated respectively and according to the product of localdensity and distance the cluster centers are determined andthe data points in the noncluster are clustered based ondensity -e nest dimension data point set represented bythe clustered characters is reordered according to the timingof the original motion sequence to obtain the behavior stringcorresponding to the original motion capture sequence so asto realize the string representation of humanmotion capturedata sequence which is shown in Figure 3

32 Motion Behavior Cluster Based on Improved Mean-ShiftAlgorithm In this paper the quotient dimension datapoints representing the original motion sequence are clus-tered by clustering -ere are many clustering methods -ereal K-means and cure clustering methods need to manuallyspecify the number of cluster centers and the method in thispaper can automatically find the number of clusters -eBIRCH clustering algorithm has good clustering perfor-mance for spherical clusters but it cannot work well forirregular clusters -e method in this paper can calculate thelocal density by using the Gaussian kernel function and itcan also cluster nonspherical clusters

Density estimation is to estimate the probability densitydistribution from a set of observations of unknown prob-ability density distribution -ere are usually two methodsparametric method and nonparametric method Given n

data points in d-dimensional Euclidean space the kerneldensity is estimated withK and the width of the window is htherefore

1113954f(x) 1

nhd

1113944

n

i1K

x minus xi

h1113874 1113875 (1)

-e mean-shift algorithm actually uses the gradientmethod to iteratively calculate the extreme points of theprobability density function According to the probabilitydensity function of the data the gradient is

nabla1113954f(x) nx

n hdcd1113872 1113873

d + 2h2

1nx

1113944xiisinSh

xi minus x( 1113857

⎧⎪⎨

⎪⎩

⎫⎪⎬

⎪⎭ (2)

Kinect 2

Client 1

SwitchServer

OSC

OSC

Kinect 1

Client 2

y1

x1

z1

y2

x2z2

yw

xwzw

Figure 2 System scheme design based on two Kinects


where the range Sh and (z) is a hypersphere with radius h andvolume hdcd and its center is x and contains nx data points

-e mean-shift vector M(x) is defined as

M(x) 1nx

1113944xiisinSh

xi minus x( 1113857 1nx

1113944xiisinSh

xi minus x (3)

-erefore formula (3) can be rewritten as

M(x) h2

d + 2nabla1113954f(x)

1113954f(x) (4)

-erefore the mean-shift vector is the difference be-tween the local mean and the center of the window and thedirection points to the peak or valley of the probabilitydensity function -e improved mean-shift algorithm isshown in Figure 4 -e implementation of the mean-shiftalgorithm is shown (a) calculate the probability distributionand similarity coefficient of the candidate target at y0according to the initial position y0 (b) calculate the weightaccording to the formula (c) calculate the new position y1according to the formula (d) calculate the probability dis-tribution and similarity coefficient of the candidate target aty1 (e) if |y1-y0 |ltC the iteration ends otherwise y0 y1skip to Step 1 and continue -e mean-shift algorithm rises(falls) at the fastest speed with variable step size and finallyconverges to the peak (Valley) point of the probabilitydensity function see Figure 4

4 Simulation and Results Analysis

41 Experimental Environment TestMethod and EquipmentIn order to verify the effectiveness of the physical trainingsystem based on virtual reality technology in this paper the

verification environment is designed as follows run thealgorithm in the software environment of MATLAB 2011the running platform is Intel Core 2 Duo the CPU is300GHz and the memory is 2GB -is paper verifies thesuperiority and reliability of the algorithm and system fromthree angles (1) capturemotion data in VR system and verifythe accuracy of motion data capture evaluation based onbehavior string (2) compared with the traditional motioncapture algorithm the superiority of motion data captureevaluation based on behavior string is verified (3) thecontrol group was set up and compared with the traditionalcontrol group to verify the superiority of the system

In addition the two technicians need to make a nor-mative judgment on the forehand and backhand strokes ofthe two groups of students under the same and relativelysimple circumstances -e final grade is the average of thescores given by the two teachers-e scoring standard is 100which is the full score in which the action integrity andaction coordination account for 50 points respectively -eevaluation of action integrity is divided into preparationposture backswing lead swing and follow-up corre-sponding to the full score of 10 15 15 and 10 respectivelythe scores of movement coordination are normative co-herence force coordination natural pace movement andstable centre of gravity with full scores of 15 15 10 and 10respectively

Test method the instructor stands two meters in front ofthe right of the tested students and throws the tennis ball tothe tested students -e student stands at the centre line ofthe field with a racket and tests 10 balls in total When thetested student tests the forehand the forehand swings andhits the ball diagonally opposite the field and the backhandneeds to swing and hit the ball diagonally opposite the field

Cluster

Sequence recovery

Sports field model

High-dimensionaldataset

Similaritycalculation

Cluster distancecalculation

BehaviorString

Human bonedata

Human bonedata

Virtual charactermodel

Figure 3 -e process of teaching mode construction based on DBN-DELM


-e two tennis technicians scored strictly according to thestandard

-e experimental equipment mainly includes a tapemeasure a whistle a stopwatch a height tester severalmarkers a weight tester a vital capacity tester 80 blowers 40tennis rackets 200 tennis balls 2 tennis courts 1 laptop 5VR equipment zero basic tennis introduction video andhighlights of tennis stars

42 -e Motion Data Capture Evaluation Based on BehaviorString -e efficiency of the big data platform is not verified-is paper selects the common Hadoop framework platformfor comparison the frameworks of Hadoop are shown inFigure 5

According to the components of elbow joints in x y andz directions their motion is periodic -e law of humanmovement is not obvious but it can be seen that the elbowjoint is doing curve movement and there are still someunsmooth and folding phenomena in the process ofmovement

-e arbitrary parameters of human motion mainly in-clude joint displacement linear velocityacceleration jointangle and angular velocityacceleration From the humanjoint point sampling data the parameters of each joint canbe obtained through calculation and analysis -e positioninformation of each joint can be obtained directly throughthe action capture equipment -e displacement of the jointcan be obtained through calculation -e change of dis-placement can be obtained through the subtraction opera-tion of the data of the first and second frames -e linearvelocity displacementdivide time can be obtained by using the

motion trajectory of the joint Acceleration reflects a changein the motion speed of human joints which can also beobtained by using the change track of human motion speed-e kinematics of the human upper limb elbow joint isanalysed and its motion parameters are obtained -emotion trajectory of the elbow joint is fitted with its regularcurve by the least square method According to the principleof forward kinematics because the wrist joint is the childnode of the elbow joint the motion of the elbow joint drivesthe motion of the elbow joint Accordingly when the wristjoint does not move the motion parameters of wrist andelbow are consistent However in the process of data

Δf(x)=xiisinSh

nx

n(hdcd)d+2h2 nx

1 (xi - x)

Moving image

Search centroid

Target center point Yes

Move window toadjust centroid

Set search area

Initialization

Histogramcalculation

Probabilitydistribution diagram


Converge

Figure 4 Implementation flow of the mean-shift algorithm

0 5x Direction

10

5

0

0 5x Direction

20

-2-4-6-8

0 5x Direction

20

-2-4-6-8

4

y Position0 5 10 15

4000

3000

2000

1000

0

z Position0 5 10 15

130

120

110

x Position0 5 10 15

24

22

2

times104

Figure 5 Motion curve based on behavior string


acquisition it cannot ensure that the wrist joint completelyfollows the movement of the elbow joint In the actualprocessing process it needs to be processed by matrixoperation

43 -e Motion Data Capture Evaluation Based on BehaviorString In order to verify the superiority of the algorithm wemainly verify it from the perspective of average error andaverage time -e results are shown in Figures 6(a) and 6(b)

-e algorithm in this paper is compared with the tra-ditional particle filter algorithm (PF) literature algorithm(VLMM) and literature algorithm (single GPDM) for 15times -e comparison diagrams of average tracking accu-racy and average time consumption are shown inFigures 6(a) and 6(b) As shown in Figure 6(a) under thesame conditions the algorithm proposed in this paper hashigher tracking accuracy than other comparison algorithmsand can well meet the accuracy requirements in 3D handtracking -e precision of the traditional PF algorithm is thesame as that of the VLMM algorithm -e single GPDMalgorithm reduces the effect of real-time human hand pa-rameters because it depends too much on the model It haslow accuracy and can only be used in the three-dimensionaltracking process with low accuracy requirements As can beseen from Figure 6(b) the traditional PF algorithm andVLMM algorithm consume a lot of time which is a greatbottleneck for the real-time tracking -e method in thispaper has been greatly improved in time consumption Inthe particle sampling stage it effectively avoids blind sam-pling realizes the purpose of reducing dimension and canfully meet the needs of interaction in terms of real timeHowever compared with the single GPDM algorithm anddata glove algorithm there is still a certain gap which is alsowhere the method in this paper needs to be furtherimproved

44 -e Training System Evaluation Based on Virtual RealityTechnology Before tennis training the physical function ofthe students participating in the experiment was evaluatedaccording to the standards of height weight 50m sprintvital capacity and standing long jump in the national stu-dent physical health standard In the tennis skill test stagethe test content includes two items forehand and backhandoblique stroke and forehand and backhand accuracy stroke

-emain research content of this paper is to test whetherthere is a significant difference in teaching effect between thenew tennis training method and the traditional trainingmode using virtual reality technology In order to eliminatethe influence of other factors on the experiment we shouldfirst test the five basic functions of the students in the ex-perimental group and the control group before the exper-iment because the experimental samples are from studentsaged 15ndash17 and the students are in the period of growth anddevelopment and it is very necessary to carry out statisticaltest and analysis-is experiment is divided into two groups20 people in each group 10 male students and 10 femalestudents -e comparison of the data in Table 1 shows thatthere is no significant difference in the physical state between

the two groups before tennis class (Pgt 005) which can beused as the sample of this experiment

40 students without tennis training foundation wererandomly selected as the research objects -ey were dividedinto two groups experimental group and control groupwith 10 male and 10 female students in each group At theend of the two courses the differences between the twogroups were analysed

Experiment 1 Ball feeling test-e ball feeling test as the name suggests is the feeling of

the ball To be exact it refers to the ability to predict the ballat the moment of hitting the ball as well as the physicalcoordination ability and feeling of the ball when hitting theball Before the experiment all students participating in theexperiment should be tested for the ball feeling It should beemphasized that the ball feeling test in the static state is donein this experiment -e ball feeling test mainly includes twoaspects the first is to test the studentsrsquo perception of the ballin motion and the second is to test the studentsrsquo sensitivityto the ball In the process of learning tennis the ball feelingand ball nature are very important -e better the studentsrsquosense of the ball the stronger their control over the ball Atthe same time the better the sense of the ball the strongertheir belief in tennis and interest will increase -ereforecultivating studentsrsquo ball sense is very important for tennisteaching -erefore in the teaching process teachers shouldpay attention to cultivating studentsrsquo ball sense and furtherenhance studentsrsquo interest in learning tennis -e quality ofball sense will also affect the progress of teaching -ereforebefore doing the experiment students should have a pro-fessional ball sense test -e results are shown in Table 2 andFigure 7

As shown in Figure 7 35 of the people in the virtualtraining experimental group have achieved excellence whileonly 10 of the people in the control group (90ndash100) haveachieved excellence and only one third of the people in thevirtual training group have achieved excellence In the goodrange (80ndash89) the number of students from both sides isbasically the same 40 in the virtual training group 35 inthe control group and 40 in the general grade range(70ndash79) while only 20 in the virtual training experimentalgroup -erefore in general most of the students in thecontrol group will get medium grades and those in the passrange and the virtual training group accounted for 5 whilethe control group accounted for 10

Experiment 2 Accuracy test of forehand and backhandstroke

After the forehand and backhand oblique stroke testfurther judge the accuracy of studentsrsquo stroke so as to betterexplore studentsrsquo mastery of tennis skills -e experimentalresults are shown in Table 3 According to Table 3 it can bejudged that the scores of the accuracy of forehand obliquestroke of the experimental group and the students of tra-ditional mode are 1128 and 1003 respectively and thedifference between the two is 125 points only the averagescore of the backhand oblique shot was 932 in the exper-imental group 818 in the traditional model control group


Table 1 -e table for the testing group

Experimental group Control groupt P

x S x SHeight (man) 17425 528 17325 461 0274 gt005Weight (man) 6636 402 6585 396 0404 gt005Height (woman) 16534 488 16427 463 0711 gt005Weight (woman) 5005 374 4976 338 1002 gt005

Table 2 -e comparison of the ball feeling test


x S x S20 s racket in place 2216 314 2296 383 minus0722 gt00520 s smash 2356 378 2192 305 1510 gt00520 s landing smash 1726 469 1687 415 0279 gt005

0 5 10 15 20number

140

130

120

110

100

90

80

70

BSPF

VLMMGPDM

aver

age a

ccur

acy

(a)

0 5 10 15 20

number

140

130

120

110

100

90

80

70

BSPF

VLMMGPDM

aver

age a

ccur

acy

(b)

Figure 6 -e comparison of motion data capture evaluation (a) Comparison of average accuracy (b) Comparison of average operationtime

90-100 80-90 70-80 60-70 lt60grade

0

5

10

15

20

25

30

35

40

45

50

prop

ortio

n (

)

the experimental groupcontrol group

Figure 7 -e score of the ball feeling test


and 114 in the difference Only the experimental resultsshow that there are significant differences between the twoteaching methods (Plt 005) -e accuracy of the studentsrsquoforward and back hand oblique stroke is higher than that ofthe traditional teaching mode -e reason may be that whenstudents carry out virtual teaching learning the classroomcontent teaching is lively and interesting and has profes-sional skills Students not only pay attention to the trainingof skills but also cultivate the interest in learning tennis -elively and interesting teaching mode makes students morefocused and conscientious which greatly mobilizes theenthusiasm and initiative of studentsrsquo learning Students invirtual tennis technology learning have greatly improved theball processing and can adjust the body and ball distance andthe degree of centre of gravity deviation according to theirown situation With the gradual improvement of proficiencyin the technology it is better and better to control themselvesand the ball

5 Conclusion

-is paper constructs the framework of sports training basedon virtual reality technology and designs a motion capturedata algorithm based on behavior string which successfullyimproves the advantages of virtual reality technology Inaddition the training experiment method is used to verifythe effectiveness and superiority of the system focusing onthe high-school students learning tennis for the first time asthe experimental object to study the differences in trainingeffects between the traditional training method and thetraining method using virtual reality technology -e testresults showed that in terms of technical skill evaluation thevirtual reality experimental group achieved excellent resultswhich was significantly different from the traditionaltraining group (Plt 005) -e average scores of forehandand backhand oblique stroke in the experimental group were1628 and 1332 respectively while the average scores of thetraditional mode training group were 1432 and 1204 re-spectively (Plt 005) in terms of the accuracy of forehandand backhand stroke the average scores of the virtual realityexperimental group were 1128 and 932 and those of thetraditional mode training group were 1003 and 818 re-spectively (Plt 005) In terms of the score of forehand andbackhand hitting skills the average scores of the subjects inthe experimental group are 8267 and 7816 and the averagescores of the students in the control group are 7511 and7135 (Plt 005) All these show that the training method ofthe experimental group is more suitable for studentsrsquo tennistraining To process virtual reality the video based on theexisting framework can be compressed and optimized Inorder to better match the processing characteristics of the

virtual reality video improve more coding efficiency andreduce more coding time and reducing code stream outputis the focus of future research

Data Availability

-e data used to support the findings of the study are in-cluded within the paper

Conflicts of Interest

-e authors declare that they have no conflicts of interest

References

[1] D -omas-Vazquez and D Singh ldquoVirtual reality in surgicaltrainingrdquo International Journal of Surgery vol 37 no 4pp 34ndash38 2021

[2] C Li and J Cui ldquoIntelligent sports training system based onartificial intelligence and big datardquo Mobile Information Sys-tems vol 2021 no 1 Article ID 9929650 11 pages 2021

[3] M Zahabi and A Razak ldquoAdaptive virtual reality-basedtraining a systematic literature review and frameworkrdquoVirtual Reality vol 24 no 1 pp 725ndash752 2020

[4] F Guo F R Yu H Zhang H Ji V C M Leung and X LildquoAn adaptive wireless virtual reality framework in futurewireless networks a distributed learning approachrdquo IEEETransactions on Vehicular Technology vol 69 no 8pp 8514ndash8528 2020

[5] L Rigamonti U-V Albrecht C Lutter M TempelB Wolfarth and D A Back ldquoPotentials of digitalization insports medicinerdquo Current Sports Medicine Reports vol 19no 4 pp 157ndash163 2020

[6] C Page P M Bernier and M Trempe ldquoUsing video sim-ulations and virtual reality to improve decision-making skillsin basketballrdquo Journal of Sports Sciences vol 37 no 3 pp 1ndash82019

[7] T R Kautzmann and P A Jaques ldquoEffects of adaptivetraining on metacognitive knowledge monitoring ability incomputer-based learningrdquo Computers amp Education vol 129pp 92ndash105 2019

[8] A Akbas W Marszałek A Kamieniarz J PolechonskiK J Słomka and G Juras ldquoApplication of virtual reality incompetitive athletes - a reviewrdquo Journal of Human Kineticsvol 69 no 1 pp 5ndash16 2019

[9] S Wei P Huang R Li Z Liu and Y Zou ldquoExploring theapplication of artificial intelligence in sports training a casestudy approachrdquo Complexity vol 2021 no 8 Article ID4658937 8 pages 2021

[10] R Lohre A J Bois J W Pollock et al ldquoEffectiveness ofimmersive virtual reality on orthopedic surgical skills andknowledge acquisition among senior surgical residentsrdquoJAMANetwork Open vol 3 no 12 Article ID e2031217 2020

[11] C Arrighi L Rossi E Trasforini et al ldquoQuantification offlood risk mitigation benefits a building-scale damage



x S x SForehand stroke 1138 174 1002 134 2546 lt005lowastBackhand stroke 932 112 830 105 334 lt005lowastlowast

Plt 005lowast means difference Plt 005lowastlowast means obvious difference


assessment through the RASOR platformrdquo Journal of Envi-ronmental Management vol 207 pp 92ndash104 2018

[12] Z Wu and C Zhou ldquoEquestrian sports posture informationdetection and information service resource aggregation sys-tem based on mobile edge computingrdquo Mobile InformationSystems vol 2021 no 3 Article ID 4741912 10 pages 2021

[13] J P Hynes J Walsh and T P Farrell ldquoRole of musculo-skeletal radiology in modern sports medicinerdquo Seminars inMusculoskeletal Radiology vol 22 no 5 pp 582ndash591 2018

[14] J Synovec L Plumblee W Barfield and H Slone ldquoOrtho-pedic in-training examination an analysis of the sportsmedicine section-an updaterdquo Journal of Surgical Educationvol 76 no 1 pp 286ndash293 2019

[15] F W Otte S K Millar and S Huttermann ldquoHow does themodern football goalkeeper train ndash an exploration of expertgoalkeeper coachesrsquo skill training approachesrdquo Journal ofSports Sciences vol 38 no 4 pp 1ndash9 2019

[16] J Pang X Li and X Zhang ldquoCoastline land use planning andbig data health sports management based on virtual realitytechnologyrdquo Arabian Journal of Geosciences vol 14 no 12pp 1ndash15 2021

[17] A Slesicka and A Kawalec ldquoAn application of the orthogonalmatching pursuit algorithm in space-time adaptive process-ingrdquo Sensors vol 20 no 12 p 3468 2020

[18] X Pan ldquoBrain-machine interface training system of motorimagery based on virtual realityrdquo NeuroQuantology vol 16no 6 pp 715ndash719 2018

[19] Y Song ldquoApplication of blockchain-based sports health datacollection system in the development of sports industryrdquoMobile Information Systems vol 2021 no 1 Article ID4663147 6 pages 2021

[20] P M Cogswell J D Trzasko and E M Gray ldquoApplication ofadaptive image receive coil technology for whole-brain im-agingrdquo American Journal of Roentgenology vol 216 no 2pp 1ndash8 2020

[21] O Besson F Vincent and S Matteoli ldquoAdaptive targetdetection in hyperspectral imaging from two sets of trainingsamples with different meansrdquo Signal Processing vol 181Article ID 107909 2021


Some researchers are committed to the application ofvirtual reality in physical training [7] Akbas et al [8] holdthat ldquothe application of virtual reality technology in physicaleducation includes five aspects knowledge learning virtualsports experiment skill training physical education networkdistance teaching and academic exchange Wei et al [9] andothers proposed that the use of virtual reality technology canmake physical education courseware and carry out physicaltraining Lohre et al [10] hold that the application of virtualreality technology in physical education includes five as-pects physical education teaching and training makingphysical education network distance teaching accidentprevention and academic exchange Arrighi et al [11] holdthat ldquothe application of virtual reality technology in collegeteaching includes virtual campus virtual classroom virtualexperiment and FGH courseware designWu and Zhou [12]put forward that ldquothere are four points in the application ofvirtual reality technology in Physical Education Teachingone is to simulate sports scenes and break through thelimitation of time and space second innovate the educa-tional model to improve studentsrsquo interest in learning andteaching effect third digital learning tracking learning dataand effectively monitoring learning behavior fourth it hasstrong functional development and broad applicationprospects in college physical education [13]rdquo Synovec et al[14] hold that the application of computer-based ldquovirtualrealityrdquo technology in college sports training has the fol-lowing four points first the comparison between sportstechnical action and reality is completed through ldquovirtualrealityrdquo technology the second is to use ldquovirtual realityrdquotechnology to realize the comparison between virtual actionand virtual reality third build different training environ-ments through ldquovirtual realityrdquo technology fourth the use ofcomputer ldquovirtual realityrdquo technology can realize remoteinteractive training [15]

-e application of virtual reality technology in collegephysical education teaching and training is the main trend ofcollege physical education development It can effectivelystimulate studentsrsquo sensory cognition and thinking modeand plays an important role in strengthening the effect ofstudentsrsquo physical education learning and training -ere-fore the contributions of this paper are as follows (1) designthe motion training framework based on virtual realitytechnology (2) design the motion data capture algorithmbased on behavior string to capture the motion trajectory ofthe target (3) the control group was designed and comparedwith the traditional training methods to verify the advan-tages of the modern training system based on VR -roughthe research and understanding of virtual reality technologyfind out the advantages of virtual reality technology andmake it flexibly applied to high-school tennis training whichcannot only enrich the training content and training meansbut also improve the training efficiency and training effect oftennis class and create a good learning atmosphere andlearning environment so as to improve the training quality-erefore this paper constructs the framework of sportstraining based on virtual reality technology and designs amotion capture data algorithm based on behavior stringwhich successfully improves the advantages of virtual reality

technology With the improvement of studentsrsquo interest inlearning the mastery of tennis skills is also improving whichis conducive to the further research of tennis training theory

2 Structure and Design of Sport TrainingBased on Virtual Reality

21 Architecture Design of Cloud Platform According to theproblems existing in the current sports training system basedon virtual reality the motion capture technology is intro-duced into the human motion posture simulation andanalysis system [16]-e data of the files obtained by motioncapture are analysed by the least square method and theresults are applied to the simulation system which will helpto solve many problems and deficiencies in the humanmotion posture simulation system [17 18] For the researchwork of using motion capture technology to obtain data andextract human motion posture parameters the developmentof this system is completed with the support of Bessel curvetheory -e physical training system based on virtual realityis mainly composed of three functional modules data ac-quisition module motion posture capture module interfacemanagementmodule and system kernel module-e overallframework of the sports training system based on virtualreality is shown in Figure 1

-e physical training system based on virtual realityadopts a hierarchical architecture which is divided into uservisualization layer signal processing layer and signal ac-quisition layer from top to bottom -e user visualizationlayer is responsible for providing a ldquocontainerrdquo for allfunctional modules receiving various messages provided byusers and transmitting the received information to the lowerlayer that is the signal processing layer so as to completethe interaction between users -e user interface module is aportable module of the presentation layer which is mainlyused to present the user interface with certain personalityand style Although the user interface module is relativelysingle and the personality is not very distinctive it has verygood interaction and powerful function and can completethe basic functions required by the system -e threemodules of model Coleman filter and human motionposture simulation in the signal processing layer belong tothe system core -e data collected by 3D suit motioncapture equipment are applied to human motion posturesimulation and combined with Bezier curve and numericaldata to complete realistic humanmotion posture simulationthe realistic simulation of the whole human motion postureis completed in these three modules the Bezier curvemodule is an extension module of the logic layer It is re-sponsible for freely editing curves to achieve the purpose ofintuitive and realistic simulation of human motion -emotion data acquisition module is combined with themotion data acquisition module of the human body modelto achieve the purpose of reading the motion data of thehuman body model

Practice has proved that with the hierarchical archi-tecture the system has clearer hierarchy lower coupling andstronger scalability -erefore the system is easy to main-tain improve and expand -e independence of modules







Head motion

MCU

Voice recognizer

Bessel curve

Computer



Kalmanfiltering

Motion capture

User visualization









1113954f(x) 1

nhd

1113944

n

i1K

x minus xi

h1113874 1113875 (1)


nabla1113954f(x) nx

n hdcd1113872 1113873

d + 2h2

1nx

1113944xiisinSh

xi minus x( 1113857

⎧⎪⎨

⎪⎩

⎫⎪⎬

⎪⎭ (2)

Kinect 2

Client 1

SwitchServer

OSC

OSC

Kinect 1

Client 2

y1

x1

z1

y2

x2z2

yw

xwzw





M(x) 1nx

1113944xiisinSh


1113944xiisinSh

xi minus x (3)


M(x) h2

d + 2nabla1113954f(x)

1113954f(x) (4)







Cluster

Sequence recovery

Sports field model




BehaviorString

Human bonedata

Human bonedata










Δf(x)=xiisinSh

nx

n(hdcd)d+2h2 nx

1 (xi - x)

Moving image

Search centroid



Set search area

Initialization




Converge


0 5x Direction

10

5

0

0 5x Direction

20

-2-4-6-8

0 5x Direction

20

-2-4-6-8

4

y Position0 5 10 15

4000

3000

2000

1000

0

z Position0 5 10 15

130

120

110

x Position0 5 10 15

24

22

2

times104






















0 5 10 15 20number

140

130

120

110

100

90

80

70

BSPF

VLMMGPDM

aver

age a

ccur

acy

(a)

0 5 10 15 20

number

140

130

120

110

100

90

80

70

BSPF

VLMMGPDM

aver

age a

ccur

acy

(b)


90-100 80-90 70-80 60-70 lt60grade

0

5

10

15

20

25

30

35

40

45

50

prop

ortio

n (

)





5 Conclusion



Data Availability




References


































Head motion

MCU

Voice recognizer

Bessel curve

Computer



Kalmanfiltering

Motion capture

User visualization









1113954f(x) 1

nhd

1113944

n

i1K

x minus xi

h1113874 1113875 (1)


nabla1113954f(x) nx

n hdcd1113872 1113873

d + 2h2

1nx

1113944xiisinSh

xi minus x( 1113857

⎧⎪⎨

⎪⎩

⎫⎪⎬

⎪⎭ (2)

Kinect 2

Client 1

SwitchServer

OSC

OSC

Kinect 1

Client 2

y1

x1

z1

y2

x2z2

yw

xwzw





M(x) 1nx

1113944xiisinSh


1113944xiisinSh

xi minus x (3)


M(x) h2

d + 2nabla1113954f(x)

1113954f(x) (4)







Cluster

Sequence recovery

Sports field model




BehaviorString

Human bonedata

Human bonedata










Δf(x)=xiisinSh

nx

n(hdcd)d+2h2 nx

1 (xi - x)

Moving image

Search centroid



Set search area

Initialization




Converge


0 5x Direction

10

5

0

0 5x Direction

20

-2-4-6-8

0 5x Direction

20

-2-4-6-8

4

y Position0 5 10 15

4000

3000

2000

1000

0

z Position0 5 10 15

130

120

110

x Position0 5 10 15

24

22

2

times104






















0 5 10 15 20number

140

130

120

110

100

90

80

70

BSPF

VLMMGPDM

aver

age a

ccur

acy

(a)

0 5 10 15 20

number

140

130

120

110

100

90

80

70

BSPF

VLMMGPDM

aver

age a

ccur

acy

(b)


90-100 80-90 70-80 60-70 lt60grade

0

5

10

15

20

25

30

35

40

45

50

prop

ortio

n (

)





5 Conclusion



Data Availability




References


































1113954f(x) 1

nhd

1113944

n

i1K

x minus xi

h1113874 1113875 (1)


nabla1113954f(x) nx

n hdcd1113872 1113873

d + 2h2

1nx

1113944xiisinSh

xi minus x( 1113857

⎧⎪⎨

⎪⎩

⎫⎪⎬

⎪⎭ (2)

Kinect 2

Client 1

SwitchServer

OSC

OSC

Kinect 1

Client 2

y1

x1

z1

y2

x2z2

yw

xwzw





M(x) 1nx

1113944xiisinSh


1113944xiisinSh

xi minus x (3)


M(x) h2

d + 2nabla1113954f(x)

1113954f(x) (4)







Cluster

Sequence recovery

Sports field model




BehaviorString

Human bonedata

Human bonedata










Δf(x)=xiisinSh

nx

n(hdcd)d+2h2 nx

1 (xi - x)

Moving image

Search centroid



Set search area

Initialization




Converge


0 5x Direction

10

5

0

0 5x Direction

20

-2-4-6-8

0 5x Direction

20

-2-4-6-8

4

y Position0 5 10 15

4000

3000

2000

1000

0

z Position0 5 10 15

130

120

110

x Position0 5 10 15

24

22

2

times104






















0 5 10 15 20number

140

130

120

110

100

90

80

70

BSPF

VLMMGPDM

aver

age a

ccur

acy

(a)

0 5 10 15 20

number

140

130

120

110

100

90

80

70

BSPF

VLMMGPDM

aver

age a

ccur

acy

(b)


90-100 80-90 70-80 60-70 lt60grade

0

5

10

15

20

25

30

35

40

45

50

prop

ortio

n (

)





5 Conclusion



Data Availability




References































M(x) 1nx

1113944xiisinSh


1113944xiisinSh

xi minus x (3)


M(x) h2

d + 2nabla1113954f(x)

1113954f(x) (4)







Cluster

Sequence recovery

Sports field model




BehaviorString

Human bonedata

Human bonedata










Δf(x)=xiisinSh

nx

n(hdcd)d+2h2 nx

1 (xi - x)

Moving image

Search centroid



Set search area

Initialization




Converge


0 5x Direction

10

5

0

0 5x Direction

20

-2-4-6-8

0 5x Direction

20

-2-4-6-8

4

y Position0 5 10 15

4000

3000

2000

1000

0

z Position0 5 10 15

130

120

110

x Position0 5 10 15

24

22

2

times104






















0 5 10 15 20number

140

130

120

110

100

90

80

70

BSPF

VLMMGPDM

aver

age a

ccur

acy

(a)

0 5 10 15 20

number

140

130

120

110

100

90

80

70

BSPF

VLMMGPDM

aver

age a

ccur

acy

(b)


90-100 80-90 70-80 60-70 lt60grade

0

5

10

15

20

25

30

35

40

45

50

prop

ortio

n (

)





5 Conclusion



Data Availability




References



































Δf(x)=xiisinSh

nx

n(hdcd)d+2h2 nx

1 (xi - x)

Moving image

Search centroid



Set search area

Initialization




Converge


0 5x Direction

10

5

0

0 5x Direction

20

-2-4-6-8

0 5x Direction

20

-2-4-6-8

4

y Position0 5 10 15

4000

3000

2000

1000

0

z Position0 5 10 15

130

120

110

x Position0 5 10 15

24

22

2

times104






















0 5 10 15 20number

140

130

120

110

100

90

80

70

BSPF

VLMMGPDM

aver

age a

ccur

acy

(a)

0 5 10 15 20

number

140

130

120

110

100

90

80

70

BSPF

VLMMGPDM

aver

age a

ccur

acy

(b)


90-100 80-90 70-80 60-70 lt60grade

0

5

10

15

20

25

30

35

40

45

50

prop

ortio

n (

)





5 Conclusion



Data Availability




References
















































0 5 10 15 20number

140

130

120

110

100

90

80

70

BSPF

VLMMGPDM

aver

age a

ccur

acy

(a)

0 5 10 15 20

number

140

130

120

110

100

90

80

70

BSPF

VLMMGPDM

aver

age a

ccur

acy

(b)


90-100 80-90 70-80 60-70 lt60grade

0

5

10

15

20

25

30

35

40

45

50

prop

ortio

n (

)





5 Conclusion



Data Availability




References






























5 Conclusion



Data Availability




References





























Application of Adaptive Virtual Reality with AI-Enabled ...

Documents

Transcript of Application of Adaptive Virtual Reality with AI-Enabled ...