Transcription

HUMOD DatabaseA versatile and open database for the investigation,modeling and simulation of human motion dynamicsDocumentationJanis Wojtusch

Documentation 1.4Janis [email protected] Universität DarmstadtDepartment of Computer ScienceSimulation, Systems Optimizationand Robotics GroupHochschulstraße 1064289 DarmstadtGermanyDecember 22, 2017

Contents1 Introduction22 Subjects23 Motion Protocol34 Measurement Setup35 Data Processing46 Data Files77 Computational Scripts9Appendix A Treadmill Dimensions12Appendix B Landmark Abbreviations12Appendix C Muscle Abbreviations14Appendix D Joint Abbreviations14Appendix E Data Structure15Appendix F References241

1 IntroductionThe HUMOD Database, derived from Human Motion Dynamics, is a versatile and open database forthe investigation, modeling and simulation of human motion dynamics with a focus on lower limbs. Thedatabase contains raw and processed biomechanical measurement data from a three-dimensional motioncapture system, an instrumented treadmill and an electromyographical measurement system for eightdifferent motion tasks performed by a female and male subject as well as anthropometric parameters forboth subjects. The quite unique combination of biomechanical measurement data with anthropometricparameters allows to create biomechanical models of the human locomotor system and to investigate andsimulate human motion dynamics including muscle driven actuation. Besides investigations in biomechanics, the database can be of value especially for the design and development of musculoskeletalhumanoid robots and for better understanding and benchmarking human-like robot locomotion. Thebiomechanical measurement data and the source code of the applied computational scripts is open andcan be obtained free of charge from the HUMOD Database humod/The HUMOD Database is made available under the Open Database License v1.0. Any rights in individualcontents of the database are licensed under the Database Contents License v1.0. The source code islicensed under the BSD 3-Clause License. Please cite the following publication, if you are using processedor raw data or computational scripts provided in the context of the HuMoD Database in your research:J. Wojtusch and O. von Stryk (2015). HuMoD - A Versatile and Open Database for the Investigation, Modeling and Simulation of Human Motion Dynamics on Actuation Level. InProceedings of the IEEE-RAS International Conference on Humanoid Robots (pp. 74 – 79).Some texts, tables and figures in this documentation are taken from or are based on this publication.2 SubjectsA healthy female and male subject performed eight different motion tasks without shoes dressed inunderwear. The subjects were given time to become familiar with the measurement setup and equipmentbefore the measurements and to rest between the trials. The measurement procedure was reviewed andapproved by the ethical review committee of Friedrich-Schiller-Universität Jena, Germany. Both subjectsprovided informed consent in accordance with the policies of the ethical review committee. Table 1 listssome details of the subjects.Table 1: Details of the female and male subject.Subject ASubject BGenderfemalemaleAge27 yrs32 yrsHeight161 cm179 cmWeight57.3 kg84.8 kgOriginCentral EuropeCentral Europeunderpants, sports braunderpantsApril 2014November 2014ClothingDate2

3 Motion ProtocolThe subjects performed eight motion tasks, partially at different speeds or under changed conditionsresulting in thirteen trials. The motion tasks cover locomotion, interaction with an object and physicalactivity representing a sample of typical repetitive tasks and goal-oriented tasks useful for biomechanicsand humanoid robotics research. These include walking, running, squatting and jumping as well asavoiding obstacles and kicking a ball. During the first and last 10 s of each trial the force plates of theinstrumented treadmill remained unloaded. Before and after performing the particular motion task, thesubject stood still on the treadmill for at least 10 s. This idle time was increased to 20 s after fast motiontasks. Details of the single trials are summarized in Table 2.Table 2: Details of the motion tasks.#DescriptionInitial idle timeTask durationFinal idle time1.1Straight walking at 1.0 ms10 s 10 s60 s10 s 10 s1.2Straight walking at 1.5 ms10 s 10 s60 s10 s 10 s1.3Straight walking at 2.0 ms10 s 10 s60 s10 s 10 s2.1Straight running at 2.0 ms10 s 10 s60 s20 s 10 s2.2Straight running at 3.0 ms10 s 10 s60 s20 s 10 s2.3Straight running at 4.0 ms10 s 10 s60 s20 s 10 s3Sideways walking at 0.5 ms10 s 10 s60 s10 s 10 s4Transition between standingand straight running at 4.0 ms 10 s 10 s112 s10 s 10 s5.1Avoiding a long box obstacle(41 20 15 cm) at 1.0 ms10 s 10 s120 s10 s 10 s5.2Avoiding a wide box obstacle(20 41 15 cm) at 1.0 ms10 s 10 s120 s10 s 10 s6Continuous squats with armsakimbo and stopped treadmill10 s 10 s40 s10 s 10 s7Kicking a soft football (20 cm,160 g) with stopped treadmill10 s 10 s100 s10 s 10 s8Continuous jumps with armsakimbo and stopped treadmill10 s 10 s20 s10 s 10 sTransition between standing and straight running comprised accelerating from 0.0 ms to 4.0 ms at 0.1 sm2 ,holding 4.0 ms for 20 s and decelerating from 4.0 ms to 0.0 ms at -0.1 sm2 . 4 Measurement SetupThe measurements were collected at the Locomotion Lab of André Seyfarth at Technische UniversitätDarmstadt, Germany. All trials were performed on the instrumented treadmill ADAL3D-WR (Tecmachine,3

Figure 1: Schematic diagram of the instrumented treadmill.France). The belt of the treadmill runs over two force plates with four single-axis force sensors (Kistler,Switzerland) that were used to measure the vertical ground reaction forces F y of the left and right foot.The two force plates are mounted on top of four multi-axis force sensors (Kistler, Switzerland) thatmeasured lateral forces F x and Fz . All forces were recorded at 1000 Hz. The software ADIMIX Walking2.0 was used to control the treadmill and store the recorded force data. Figure 1 shows a schematicdiagram of the instrumented treadmill and the single- and multi-axis force sensors. Detailed dimensionsof the instrumented treadmill are provided in Appendix A.The motion of upper and lower limbs was recorded at 500 Hz with a three-dimensional motion capturesystem consisting of four Oqus 310 cameras and six Oqus 300 cameras (Qualisys, Sweden). A set ofthirty-five reflective markers with a diameter of 19 mm mounted on thin cardboard was placed on theskin at anatomical landmarks by an experienced examiner. One additional reflective marker was placedon top of the underpants above pubic symphysis [Reed1999]. Figure 2a illustrates the locations of thethirty-six reflective markers. A description of the associated landmarks and used abbreviations is givenin Appendix B. For calibrating and controlling the motion capture system, storing the recorded motiondata as well as assigning the recorded motion data to the individual markers, the software TrackManager2.7 was applied.The electrical activity of fourteen selected skeletal muscles in the legs was recorded at 2000 Hz with theelectromyographical measurement system Bagnoli-16 Desktop (Delsys, USA). The measured signals wereinternally filtered to a bandwidth between 20 Hz and 450 Hz. The set of fourteen surface electrodes wasplaced according to SENIAM guidelines [Hermens2000] by an experienced examiner. Figure 2b showsthe locations of the fourteen surface electrodes for electromyographical measurement. The associatedmuscles and used abbreviations are listed in Appendix C. The software EMGworks Acquisition 3.6 wasused to control the electromyographical measurement system and store the recorded activity data.5 Data ProcessingThe raw data measured with the motion capture system, instrumented treadmill and electromyographical measurement system was exported into the MAT file format and processed with the numerical computing software MATLAB (MathWorks, USA) in order to provide additional information for the investigation, modeling and simulation of human motion dynamics.4

(a)(b)Figure 2: Locations of the thirty-six reflective markers for motion capture in (a) and the fourteen electrodes for electromyographical measurement in (b).Raw motion and force data was synchronized by compensating temporal offset and drift as well as transforming the global reference frame of the motion capture system into the global reference frame of theinstrumented treadmill considering the ISB recommendations for reference frame notation [Wu1995].Figure 1 illustrates the applied global reference frame where the origin is located at the center of therectangle spanned by the left and right force plates projected to the top of the belt surface.Infrequent gaps in the raw kinematic motion data of up to 300 ms resulting from temporarily coveredreflective markers were filled by applying polynomial approximations. The measured spatial positionsof the reflective markers were then shifted to the approximated skin surface. This was achieved by approximating a normal vector perpendicular to the skin surface pointing towards the considered reflectivemarker from adjacent reflective markers and estimated joint centers. The normalized normal vector wasmultiplied with the reflective marker radius and additional support material thickness and subtractedfrom the measured spatial position.GLAThe normal vector is parallel to the line connecting the midpoint between the TRAmarkers with the GLA marker.TRAThe normal vector is parallel to the line connecting the TRA markers.SUP, C7The normal vector is parallel to the line connecting the C7 and SUP markers.T8The normal vector is parallel to vector sum of the normal vectors specified for theSUP, C7 and T12 markers.T12The normal vector is parallel to the line connecting the T8 and T12 markers rotatedby π2 rad about the line connecting the ACR markers.ACRThe normal vector is perpendicular to the normal vector specified for the SUP andC7 markers and the line connecting the ACR markers.LHCThe normal vector is perpendicular to the lines connecting the WRI and LHC markers as well as the estimated shoulder joint center and LHC marker.5

ASIS, PSISThe normal vector is parallel to the line connecting the midpoint between the ASISmarkers with the midpoint between the PSIS markers.PSThe normal vector is parallel to the line connecting the midpoint between the PSISmarkers with the PS marker.GTRThe normal vector is parallel to the line connecting the GTR markers.LFC, MFCThe normal vector is parallel to the line connecting the LFC and MFC markers.LM, MMThe normal vector is parallel to the line connecting the LM and MM markers.CALThe normal vector is parallel to the line connecting the CAL and MT2 markersMT2, MT5, HALThe normal vector is perpendicular to the lines connecting the CAL and MT5 markers as well as the MT2 and MT5 markers.The shifted spatial positions of the reflective markers were then used to estimate the joint centers offifteen joints in arms, trunk, pelvis and legs by applying established regressing equations. A descriptionof the used abbreviations is given in Appendix D.LNJThe lower neck joint center was estimated from the C7, SUP and ACR markersaccording to Reed et al. [Reed1999].SJL , SJRThe shoulder joint centers were estimated from the C7, SUP and ACR markersaccording to Reed et al. [Reed1999].EJL , EJRThe elbow joint centers were estimated from the WRI and LHC markers as well asthe estimated shoulder joint centers according to Reed et al. [Reed1999].ULJThe upper lumbar joint center was estimated from the C7, T8, T12, SUP and ACRmarkers according to Reed et al. [Reed1999] and Dumas et al. [Dumas2015].LLJThe lower lumbar joint center was estimated from the ASIS, PSIS and PS markersaccording to Reed et al. [Reed1999].HJL , HJRThe hip joint centers were estimated from the ASIS, PSIS and PS markers accordingto Harrington et al. [Harrington2007].KJL , KJRThe knee joint centers were estimated from the LFC and MFC markers accordingto Dumas et al. [Dumas2007a].AJL , AJRThe ankle joint centers were estimated from the LM and MM markers accordingto Dumas et al. [Dumas2007a].TJL , TJRThe toe joint centers were estimated from the CAL, MT2 and MT5 markers basedon definitions by Zatsiorsky [Zatsiorsky1998].Additional regression equations from literature for hip, knee and ankle joints were implemented andcan be used alternatively by applying the provided computational scripts [Reed1999; Leardini1999;Seidel1995; Davis1991; Bell1990; Dempster1955; Hicks1953].For the estimation of the joint trajectories including joint positions, velocities and accelerations, a Kalmansmoother in combination with a subject-specific forward kinematics model with thirty degrees of freedomand fourteen body segments was applied [DeGroote2008; Yu2004]. This approach allows to reduce theinfluences of instrumental errors and soft tissue artifacts. The forward kinematics model consists of ahead, thorax and abdomen segment, two upper and lower arm segments, a pelvis segment and twothigh, shank and foot segments. The model structure is shown in Figure 3. The joint trajectories aregiven as Tait–Bryan angles in x - y 0 -z 00 convention.6

Figure 3: Forward kinematics model with thirty degrees of freedom.The raw ground reaction force data was filtered using a sixth order zero-lag low-pass filter with a cut-offfrequency of 50 Hz. In order to decompose the measured lateral ground reaction forces F x and Fz andthe measured vertical ground reaction force F y in the event of mixed force plate contact during doublesupport phase for the locomotion trials, parametrized transition functions determined using a multipleregression analysis were applied [Villeger2014]. The transition functions approximate the ground reaction force decrease of the foot leaving the ground during double support phase. The ground reactionforce data was then used to estimate the center of pressure and detect individual events like left andright steps, squats or kicks.The raw muscle activity data was rectified and filtered using a root-mean square filter with a window sizeof 300 ms [Konrad2005]. In addition, the filtered muscle activity data was normalized to the maximumactivity level over all trials of the subject. Each dataset provides filtered and non-normalized as well asfiltered and normalized muscle activity data.Subject-specific anthropometric parameters including body segment masses, centers of mass as well asmoments and products of inertia were estimated with linear regression equations [Dumas2007a; Dumas2007b; Dumas2015]. Raphaël Dumas kindly provided an updated version of the applied regressiontables with some corrections in the foot parameters. The required body segment lengths were obtainedfrom averaged kinematic motion data taken at the beginning of the trials with stopped treadmill. Theapplied joint axes match the axes of the estimated body segment inertial parameters and comply withthe ISB recommendations [Wu2002; Wu2005].6 Data FilesThe HUMOD Database website provides a number of data files that contain the raw and processed biomechanical measurement data for the different motion tasks, the anthropometric parameters for the subjectsand supplemental data. Raw and processed biomechanical measurement data as well as anthropometric7

parameters are stored in the MAT file format of the numerical computing software MATLAB (MathWorks,USA). Supplemental data is provided as PNG image files or WEBM video files. Links to the individualdata files are organized in separate tables for each subject. The following graph gives a brief overviewof the data file structure and content.Data filesSubject parameters (Parameters.mat)The Parameters.mat data file provides anthropometric parameters and meta data for the subject.Processed dataDataset (#.mat)The #.mat data files, where # stands for the number of the motion task as given in Table 2, provide the processedbiomechanical measurement data from the three-dimensional motion capture system, instrumented treadmill andelectromyographical measurement system.Raw dataMotion (#-RawMotion.mat)The #-RawMotion.mat data files, where # stands for the number of the motion task as given in Table 2, containthe raw marker coordinates measured with the three-dimensional motion capture system.Muscle (#-RawMuscle.mat)The #-RawMuscle.mat data files, where # stands for the number of the motion task as given in Table 2, containthe raw muscle activity data measured with the electromyographical measurement system.Force (#-RawForce.mat)The #-RawForce.mat data files, where # stands for the number of the motion task as given in Table 2, contain theraw ground reaction forces measured with the instrumented treadmill.Ground reference (GroundReference.mat)The GroundReference.mat data file contains reference coordinates of the instrumented treadmill that are used tomatch the global reference frames of the motion capture system and the instrumented treadmill.DiagramsMuscle (#.png)The supplemental #.png image files, where # stands for the number of the motion task as given in Table 2,visualize the processed muscle activities.Force (#.png)The supplemental #.png image files, where # stands for the number of the motion task as given in Table 2,visualize the processed ground reaction forces.VideosMotion (#.webm)The supplemental #.webm video files, where # stands for the number of the motion task as given in Table 2, playan animated visualization of the processed marker and joint center coordinates.For most applications dealing with modeling and simulation of human motion dynamics, it is sufficient to employ the subject parameter file (Parameters.mat) and the processed data files (#.mat).The subject parameters can be used to create a subject-specific biomechanical kinematics and dynamicsmodel, while the processed data files provide task-specific joint trajectories, ground reaction forces andmuscle activities. For further investigations, the raw data files (#-RawMotion.mat, #-RawMuscle.mat,#-RawForce.mat) and the ground reference file (GroundReference.mat) allow to validate or modifythe applied data processing and to derive additional information. A detailed description of the data8

file structure and content is given in Appendix E. All abbreviations used in the data files are listed inAppendices B, C and D.7 Computational ScriptsThe source code of the applied computational scripts is available in an online repository with distributedrevision control. Additional helper scripts are located in the Scripts subdirectory. When starting withthe raw biomechanical measurement data, some of the scripts require data extracted or generated by adifferent script. The following graph gives a brief overview of the computational scripts and provides asuggested execution sequence.Computational scriptsDirectory structure generation (DirectoryStructureGeneration.m)This script generates a directory structure that is used by the computational scripts. Please modify the global path inScripts/getPath.m and the local paths in Scripts/getFile.m if required.Ground reference estimation (GroundReferenceEstimation.m)This script estimates the rotation and translation parameters to transform points from the reference frame of themotion capture system into the the reference frame of the instrumented treadmill. The reference frame of the instrumented treadmill is the global reference frame for all datasets. This script creates the ground structure in theprocessed data files #.mat.Motion gap filling (MotionGapFilling.m)This tool processes the raw marker trajectories of the motion capture system and can be used to fill small gaps. It hasa graphical user interface and provides different methods for gap filling. It transforms the reference frame accordingto ISB recommendations [Wu1995] and creates the initial motion structure in the processed data files #.mat. Figure 4shows the graphical user interface and exemplary settings for filling a gap with the constrained fit method.Motion transformation (MotionTransformation.m)This script transforms the marker coordinates in the motion variable in the processed data files #.mat into the thereference frame of the instrumented treadmill. The reference frame of the instrumented treadmill is the globalreference frame for all datasets.Joint center estimation (JointCenterEstimation.m)This script estimates the marker coordinates shifted to skin surface and the joint center positions from measured andestimated marker coordinates according to predictive methods given in different references.Motion visualization (MotionVisualization.m)This script creates an animated visualization of the processed marker and estimated joint center positions.Subject parameter estimation (SubjectParameterEstimation.m)This script estimates subject parameters based on segment lengths and on regression equations [Dumas2007a; Dumas2007b; Dumas2015] and with local reference frames according to ISB recommendations [Wu2002; Wu2005].Joint trajectory estimation (JointTrajectoryEstimation.m)This script estimates the joint trajectories including joint positions, velocities and accelerations and smoothes theestimated joint center positions by applying a Kalman smoother [DeGroote2008; Yu2004] and a subject-specificforward kinematics model.Trajectory visualization (TrajectoryVisualization.m)This script creates an animated visualization of the processed joint trajectories and smoothed joint center positions.9

Force filtering (ForceFilter.m)This script processes the raw ground reaction forces of the instrumented treadmill and transforms the reference frameaccording to ISB recommendations [Wu1995]. It synchronizes motion and force data by compensating the time delaybetween the motion capture system and the instrumented treadmill. This script creates the initial force structure inthe processed data files #.mat.Force separation (ForceSeparation.m)This script smooths the measured ground reaction forces and separates the forces for left and right side by applyingparametrized transition functions [Villeger2014].Force matching (ForceMatching.m)This script matches the ground reaction forces for left and right side by shifting residual ground reaction forces duringsingle support.Force visualization (ForceVisualization.m)This script creates a visualization of the total and separate processed ground reaction forces.Event detection (EventDetection.m)This tool applies an event detection algorithm to find the start and end of events like steps or jumps. A graphical userinterface allows to check and correct the automatically detected events. For some motion tasks, the events have to bedefined manually within the graphical user interface. The tool creates the initial events structure in the processeddata files #.mat. Figure 5 shows the graphical user interface with a detected event for slow straight walking.Event visualization (EventVisualization.m)This script creates a visualization of the processed ground reaction forces with an overlay of the detected events.Center of pressure estimation (CenterOfPressureEstimation.m)This script estimates the center of pressure positions from the processed ground reaction forces, measured force sensordata and given force sensor positions. The estimated center of pressure positions are limited to the foot dimensionsin order to compensate high error amplification at low force sensor values.Muscle filtering (MuscleFilter.m)This script processes the raw muscle activities of the electromyographical measurement system. It rectifies the signalsand applies a zero-phase low-pass, moving-average or root-mean-squares filter with adjustable parameters [Konrad2005]. The script creates the initial muscle structure in the processed data files #.mat.Muscle normalization (MuscleNormalization.m)This script normalizes the filtered muscle activities by finding the global maximum values in all datasets and correctingscattered outliers.Muscle visualization (MuscleVisualization.m)This script creates a visualization of the filtered or normalized muscle activities.Meta data generation (MetaDataGeneration.m)This scripts adds some meta data to the datasets. It creates the meta structure in the processed data files #.mat.10

Figure 4: Graphical user interface of MotionGapFilling.m.Figure 5: Graphical user interface of EventDetection.m.11

A Treadmill DimensionsThe eight single-axis force sensors L1 to L4 and R1 to R4 were used to measure the vertical groundreaction forces F y of the left and right foot. The four multi-axis force sensors S1 to S4 measured thelateral forces F x and Fz .B Landmark Abbreviations12GLAGlabella: Undepressed skin surface point obtained by palpating the mostforward projection of the forehead in the midline at the level of the browridges [Reed1999].TRAL , TRARLeft and right tragion: Undepressed skin surface point obtained by palpating themost anterior margin of the cartilaginous notch just superior to the tragus of theear located at the upper edge of the external auditory meatus [Reed1999].SUPSuprasternale: Undepressed skin surface point at the superior margin of the jugular notch of the manubrium on the midline of the sternum [Reed1999].C77th cervical vertebra: Depressed skin surface point at the most posterior aspectof the spinous process of the 7th cervical vertebra [Reed1999].T88th thoracic vertebra: Depressed skin surface point at the most posterior aspectof the spinous process of the 8th thoracic vertebra [Reed1999].

T1212th thoracic vertebra: Depressed skin surface point at the most posterior aspectof the spinous process of the 12th thoracic vertebra [Reed1999].ACRL , ACRRLeft and right acromion: Undepressed skin surface point obtained by palpating the most anterior portion of the lateral margin of the acromial process of thescapula [Reed1999].LHCL , LHCRLeft and right lateral humeral epicondyle: Undepressed skin surface point atthe most lateral aspect of the humeral epicondyle [Reed1999].WRIL , WRIRLeft and right wrist: Undepressed skin surface point on the dorsal surface of thewrist midway between the radial and ulnar styloid processes [Reed1999].ASISL , ASISRLeft and right anterior-superior iliac spine: Depressed skin surface point atthe anterior-superior iliac spine. Located by palpating proximally on the midlineof the anterior thigh surface until the anterior prominence of the iliac spine isreached [Reed1999].PSISL , PSISRLeft and right posterior-superior iliac spine: Depressed skin surface point at theposterior-superior iliac spine. This landmark is located by palpating posteriorlyalong the margin of the iliac spine until the most posterior prominence is located,adjacent to the sacrum [Reed1999].PSPubic symphysis: Depressed skin surface point at the anterior margin of pubicsymphysis, located by the subject by palpating inferiorly on the midline of theabdomen until reaching the pubis. The subject is instructed to rock his or herfingers around the lower margin of the symphysis to locate the most anteriorpoint [Reed1999].GTRL , GTRRLeft and right greater trochanter: Undepressed skin surface point at the mostlateral prominent of the upper femur.LFCL , LFCRLeft and right lateral femoral epicondyle: Undepressed skin surface point at themost lateral aspect of the lateral femoral epicondyle [Reed1999].MFCL , MFCRLeft and right medial femoral epicondyle: Undepressed skin surface point at themost medial aspect of the medial femoral epicondyle.LML , LMRLeft and right lateral malleoius: Undepressed skin surface point at the mostlateral aspect of the malleolus of the fibula [Reed1999].MML , MMRLeft and right medial malleoius: Undepressed skin surface point at the mostmedial aspect of the malleolus of the tibia.CALL , CALRLeft and right calcaneus: Undepressed skin surface point at the most posteriorprominent of the calcaneus.MT2L , MT2RLeft and right 2nd metatarsal head: Undepressed skin surface point above thedistal head of the 2nd metatarsal.MT5L , MT5RLeft and right 5th metatarsal head: Undepressed skin surface point above thedistal head of the 5th metatarsal.HALL , HALRLeft and right hallux: The anterior point of the 1st digit of each foot.In the provided data files, all landmarks are identified by labels of the form [xxx][ R/ L]. The firstpart xxx is the two-, three- or four-letter landmark abbreviation as listed above. The last part L or Rindicates the left or right body side if applicable.13

C Muscle AbbreviationsSOLL , SOLRLeft and right soleus muscle: Plantar flexion of the ankle joint [Hermens2000].TIAL , TIARLeft and right tibialis anterior muscle: Dorsiflexion of the ankle joint and assistance in inversion of the foot [Hermens2000].GLSL , GLSRLeft and right gastrocnemius lateralis muscle: Flexion of the ankle joint andassist in flexion of the knee joint [Hermens2000].VSLL , VSLRLeft and right vastus lateralis muscle: Extension of the knee joint [Hermens2000].RCFL , RCFRLeft and right rectus femoris muscle: Extension of the knee joint and flexion ofthe hip joint [Hermens2000].BCFL , BCFRLeft and right biceps femoris muscle: Flexion and lateral rotation of the kneejoint. The long head also extends and assists in lateral rotation of the hipjoint [Hermens2000].GLXL , GLXRLeft and right gluteus maximus muscle: Extends, laterally rotates and lowerfibres assist in adduction of the hip joint. The upper fibres assist in adduction.Through its insertion into the iliotibial tract, helps to stabilise the knee in extension [Hermens2000].In the provided data files, all muscles are identified by labels of the form [xxx][ R/ L]. The first partxxx is the three-letter muscle abbreviation as listed above. The last part L or R indicates the left orright body side if applicable.D Joint AbbreviationsBJBase joint that connects the human model at the lower lum

LNJ The lower neck joint center was estimated from the C7, SUP and ACR markers according to Reed et al. [Reed1999]. SJ L, SJ R The shoulder joint centers were estimated from the C7, SUP and ACR markers according to Reed et al. [Reed1999]. EJ L, EJ R The elbow joint centers were estimated from the WRI and LHC markers as well as