A Public Dataset of Overground and Treadmill Walking in Healthy

Individuals Captured by Wearable IMU and sEMG Sensors

Harald Loose, Laura Tetzlaff and Jon Lindström Bolmgren

Technische Hochschule Brandenburg, University of Applied Sciences,

Magdeburger Str. 50, 14770 Brandenburg, Germany

Keywords: Public Dataset, Overground and Treadmill Walking, Wearables, IMU and sEMG Sensors.

Abstract: The paper presents our public Gait Analysis Data Base (http://gaitanalysis.th-brandenburg.de), which contains

3D walking kinematics and muscle activity data from healthy adults walking at normal, slow or fast pace on

the flat ground or at an incremental speeds on treadmill. The acceleration, angular velocity and magnetic rate

vectors are measured using XSens MTw sensors attached to both feet, shanks, thighs and the pelvis. EMG

recordings are acquired using PLUX sEMG sensors applied at various leg muscles. The paper gives not only

a detailed description of the data base, its webpage and the used terms (scenario, proband, experiment and

trial), but also an overview about the experimental setup, the acquisition of data and the procedure of the

experiments, the data processing and evaluation. Results of exemplary applications are described in the second

part of the paper. Here the focus is set on the performance of walking: the individual ability to control, to

repeat and to reproduce the pace or the dependence of gait parameters on the pre-set velocity.

1 INTRODUCTION

Since 2000 the platform PhysioNet

offers free web

access to large collections of recorded physiologic

signals (PhysioBank) and related open-source

software (PhysioToolkit). PhysioNet is an online

forum for the dissemination and exchange of recorded

biomedical signals and open-source software for

analyzing them. It provides facilities for the

cooperative analysis of data and the evaluation of

proposed new algorithms (Goldberger et al., 2000).

Gait databases

contain stride interval (gait cycle

duration) time series in text form.

Data sharing and increased acceptance of

replication studies have been advocated to overcome

experimental limitations and to validate the

inferences made by previous gait studies (Ferber et

al., 2016; Knudson, 2017). Only a handful of walking

biomechanics datasets have been made publicly

available (Hnat, Moore & Van den Bogert, 2015;

Kirtley, 2014; Willson & Kernozek, 2014). There is

an objective need to share data and to have normative

databases to improve and to evaluate the

http://www.physionet.org

https://physionet.org/physiobank /database/#gait

http://demotu.org/datasets/walk/

interpretation of gait analysis outcomes. In the 1990s

the first gait datasets were made available (Winter,

1993; Winter 2009; Perry, 1992), including patterns

for joint angles, joint moments and reaction forces,

later patterns of muscle activities were added (Bovi,

2011).

The MMClab

of the University of ABC, Brazil,

provides a public dataset of 3D walking kinematics

and kinetics data on healthy young and older adults at

a range of gait speeds in both the treadmill and

overground environments. The datasets include both

raw and processed kinematic and kinetic data

(Fukuchi et al. 2018).

Our GaitAnalysisDataBase

contains 3D walking

kinematics and muscle activity data from healthy

adults walking on the flat ground or at a treadmill.

The acceleration, angular velocity and magnetic rate

vectors are measured using inertial measurement

units (IMU Xsens MTw

) applied to both feet, shanks,

thighs and the pelvis. EMG recordings are acquired

using acceleration and surface EMG sensors (PLUX

XYZ and PLUX

sEMG) applied at various leg

muscles. The data sets include unfiltered, gravity

compensated kinematic data of Xsens sensors and

http://gaitanalysis. th-brandenburg.de

xsens.com

plux.info

164

Loose, H., Tetzlaff, L. and Bolmgren, J.

A Public Dataset of Overground and Treadmill Walking in Healthy Individuals Captured by Wearable IMU and sEMG Sensors.

DOI: 10.5220/0008879501640171

In Proceedings of the 13th International Joint Conference on Biomedical Engineering Systems and Technologies (BIOSTEC 2020) - Volume 4: BIOSIGNALS, pages 164-171

ISBN: 978-989-758-398-8; ISSN: 2184-4305

unprocessed raw data of PLUX acceleration and

sEMG sensors.

The GaitAnalysisDataBase collects data sets

acquired under the supervision of the authors during

the last years at the Brandenburg University of

Applied Sciences (THB), at FH Vorarlberg (Austria),

at the University of Oulu (Finland) and at MMUST

(Kenya), indoor and outdoor, on paved and unpaved

trails, at various climatic conditions, investigating

various aspects of human movement.

Students and professors, technicians and

researchers have been involved in the process of

preparing and executing measurements as well as in

storing and evaluating the acquired data. All these

volunteers – healthy adults between 18 and 65 from

several nationalities – provided informed consent

about the experiments, data storage and the future use

of data.

Following the initial idea of the PhysioNet

platform, the GaitAnalysisDataBase are meant to

facilitate “the cooperative analysis of data and the

evaluation of proposed new algorithms”, to support

the development of robust algorithms and to be used

for teaching and other educational purposes. The data

were collected by lecturers and students guided by

prescribed procedures and checklists. Recordings

containing measurement errors or procedural faults,

caused by equipment, subjects or instructors, have not

been excluded. They should be detected by

automatically proceeding and robustness algorithms.

The data sets serve as useful examples for testing

newly implemented algorithms.

The paper is divided into two main parts: the

description of the public dataset and exemplary

applications to demonstrate possible investigations

using the provided data. This paper is not intended to

explain all aspects of the developed

methods/algorithms, including all related to them

problems, and to discuss the results of the exemplary

applications. More details are presented in e.g. Loose

(2015) and Loose et al. (2016).

2 SYSTEM AND EXPERIMENTS

We have focused on human walking, tried to

understand the underlying process and to find the best

positions of sensors. Robust and reliable algorithms

which apply to a wide range of walking scenarios (~2-

8 km/h) were developed. The algorithms were

evaluated on data sets acquired from IMUs attached

to the foot, shank, thigh, pelvis or from sEMG,

http://www.seniam.org/

applied to various muscles, supplemented by an

accelerometer placed at the heel. Two main scenarios

- repetitive walking on the flat ground and walking on

a treadmill - were addressed with a large number of

healthy subjects.

2.1 Experimental Setup

The Xsens sensors are clipped on body straps

attached similarly on the left and right lower limbs

and one in the middle of the back. Typically one pair

is sitting on the metatarsus, two directly above the

ankle and the knee. The distances of the sensors from

the floor, as well as the length of the limbs are stored

in the subject’s individual experiment record.

Figure 1: Experimental setup – Xsens and PLUX sensors

are applied.

When PLUX sensors are included in the

experimental setup, sEMG pads are (mostly)

positioned accordingly the recommendations of

SENIAM

symmetrically on muscles involved in

locomotion activities,: m. gluteus maximus, m. rectus

femoris, m. biceps femoris, m. vastus lateralis

femoris, m. vastus medialis femoris, m. tibialis

anterior, m. tibialis posterior, m. gastrocnemius

lateralis, m. gastrocnemius medialis, m. soleus.

Accelerometers are applied to the heel to supplement

sEMG records.

A Public Dataset of Overground and Treadmill Walking in Healthy Individuals Captured by Wearable IMU and sEMG Sensors

165

2.2 Sensors and Software

During the gait analysis courses three generations of

the 9DOF Xsens MTw sensors have been used to

acquire accelerations, angular velocities and

magnetic rates as well as the sensor orientation at a

sampling rate between 60 and 100 Hz. Two

generations of the PLUX equipment have been

applied to record muscle activities (sEMG) and the

related acceleration at a sampling rate of 1 kHz.

2.2.1 Xsens MTw Sensors

The 9DOF Xsens MTw sensor (Roetenberg, 2009)

incorporates three microelectromechanical sensors:

triple-axis gyroscope, accelerometer and,

magnetometer.

Onboard the data of the primary sensors are

sampled with 1800 Hz, strapped down by integration

(SDI) incorporating the estimate of orientation to the

transfer rate 60 Hz for seven sensors or 100 Hz for

two sensors (first generation).

The MTw are connected via Bluetooth to one

Awinda station and the data acquisition software “MT

Manager, versions 3.81, 4.21 and 4.9”.

All involved sensors are synchronized with high

accuracy (< 0.01 ms). The software provides linear

acceleration a, angular velocity ω, magnetic field m

and quaternion q (estimated on-board with < 1° of

static and 2° RMS of dynamic accuracy).

Before the measurement the sensors need a calm

or slow motion for calibration, to determine the initial

orientation of the sensor with respect to the world

coordinate system.

Figure 2: IMU Xsens Mtw sensors and AVIRA unit (left)

and PLUX Channel hub and connected sensors (right).

2.2.2 PLUX Sensors

The PLUX biosignal kit

includes a wireless 8

channel hub, various sensors and the data acquisition

software "OpenSignals". The non-filtered sensor data

were acquired at a sampling rate of 1 kHz.

https://plux.info/12-biosignalsplux

In most of the experiments, 5 sEMG and a triaxial

accelerometer sensors were connected to a hub (one

for each leg), sometimes two similar hubs were used

simultaneously.

2.3 Scenarios

All data sets have been acquired in two main

scenarios:

 “The Catwalk”: walking a distance of mostly 20

m (in a range of 10 m to 80 m) on flat ground at

usual/normal, reduced/slowed and increased/fast

speed,

 “The Treadmill”: walking on a treadmill at

incremental speed settings from 3.5 to 6.5 km/h or

2 to 8 km/h).

The specifications of the experimental setup for

each dataset is given on the GaitAnalysisDataBase

website and is included in META data of any record.

A test scenario was involved in the examination

of distance accuracy: Straight forward, steady

walking outdoor on a long enough distance (~175 m),

on flat and paved ground. The distance was measured

alternatively with GPS and tape line. The trial was

repeated twice.

2.4 Cohorts

Study participants included 108 healthy volunteers

between 17 - (average: 26, median: 23) - 63 years old,

147 – (174) – 194 cm height, 54 – (76) – 120 kg

weight and a body mass index of 19 – (24) – 37. There

were 45 females and 63 males. Figure 3 shows the

distribution of age, height, and weight of all

volunteers. The 0 kg bin is caused in the lack of data

in one scenario.

Figure 3: Age, height and weight distribution.

Only in 2017 (March – October) a subject was

monitored during the recovery process after the

surgery caused by a fibula fracture (“Cat walking”

and “Treadmill).

BIOSIGNALS 2020 - 13th International Conference on Bio-inspired Systems and Signal Processing

166

2.5 Procedure

All gait trials were performed in everyday conditions,

and the participants were comfortably dressed. Before

data collection, each participant received a brief

explanation of the study and signed the consent form.

Body height and body mass were measured. The year

of birth was asked.

Each recording was started and terminated with a

three second calm phase independent of whether the

subject was walking on flat ground or on the

treadmill. Each trial is recorded and stored separately.

The subject is asked to stop at the end of a trial and

not to turn or to move before the recording is

terminated.

2.6 Evaluation

All algorithms used in all steps of data evaluation

were developed, implemented and tested in

MATLAB®

by ourselves. An open script is

organized to process experimental data automatically

step by step. After each step, the intermediate results

are saved.

The developed algorithms are designed on a

robust stride detection and a stride-by-stride

determination of derived characteristics allowing

recalibrations at the beginning of each stride.

The term “stride” or “gait cycle” is defined as the

period between two successional gait events. Perry

(Perry, 1992) and others defined the initial contact,

i.e. the moment of the heel strike, as the beginning of

the gait cycle. To choose the period from mid-stance

to mid-stance for calculation is more constructive,

because at this moment obviously the foot is not

moving, it stands on the floor and the leg is vertical

elongated.

2.6.1 Xsens Sensor Data Processing

9DOF Xsens MTw sensors provide accelerations,

angular velocities, magnetic rates and orientation

data. The sampled data of all involved in the trial

sensors are received by the Xsens MT manger via

Bluetooth and stored in one proprietary file. This data

file is exported in text csv-files, one for each sensor.

The data of each Xsens sensors are evaluated

separately. The following steps are included:

 Pre-processing: reading and reorganizing the

acquired data, given in the sensor related

coordinate system (SCS), transformation of

sensor data into world coordinate system (WCS),

www.mathworks.com

elimination of gravity, calculation of orientation

relative to the initial one, calculation of angles

between z-axes of a sensor and the vertical or the

horizontal plane, calculation of joint angles.

 Processing: estimation of direction of motion,

calculation of candidates for gait events,

plausibility check, determination of gait cycles,

transformation of data into motion coordinate

system (MCS), integration of acceleration,

calculation of velocity and position data stride by

stride.

 Post-processing: calculation of stride related and

average features, determination of average

motion.

 Evaluation: building figures, extracting and

processing tables.

The database contains the pre-processed datasets

of each sensor. Additional the following computed

parameters are available at the website: duration in

[s], distance in [m], cadence in [steps/min] and

number of strides. Other features are determined

during post-processing and evaluation.

Remark: All sensors are treated in the same way

independent on their position. Because of this the

calculated velocities do not include the steady part of

motion, which is zero for the foot sensors. The steady

motion increases together with the height of the

sensor position (see Loose, 2015).

2.6.2 PLUX Sensor Data Evaluation

sEMG sensors were mostly used in combination with

XYZ accelerometers placed at the heel, sometimes

without them. The following two different

approaches to detect gait cycles were implemented. If

an XYZ accelerometer is used the outstanding local

minima related to the heel strike are determined and

the signals are partitioned beginning and terminating

at heel strikes. Otherwise the main frequency of all

available signals is calculated. The correspondent

period of time (average stride duration) is taken to

partition them.

The following steps are included:

 Pre-processing: reading, reorganizing and

conditioning the acquired data, calculation of

sEMG envelopes based on rms-methods.

 Processing: determination of gait cycles,

partitioning sensor signals and envelopes.

 Post-processing: calculation of mean and median

frequencies, power spectrums and energy

characteristics, determination of average curves.

A Public Dataset of Overground and Treadmill Walking in Healthy Individuals Captured by Wearable IMU and sEMG Sensors

167

 Evaluation: building figures, extracting and

processing tables.

3 ORGANIZATION OF THE

DATA BASE

The data sets collected in the database were acquired

by lecturers and students, in undergraduate and

graduate courses dealing with various aspects of gait

analysis, experimental and evaluation methods. In

each course the scenario, the experimental setup and

the procedure of the trials were explained and

discussed with all participants. The execution of the

experiments following pre-defined checklists has

been supervised by the lecturer (authors). Any

measurement errors or procedural faults, caused by

equipment, probands or instructors, were noted and

were excluded only in the case that nothing

substantial was recorded. They serve as useful

examples for testing the robustness of algorithm

implementations.

Before getting access to the database, the user

needs to register in the system entering his username,

password and email address. Later he can sign in by

username and password. The personal data of the

registration are used only for statistics about access to

the database.

3.1 Terms

The most important terms in the database are

scenarios, probands, experiments, the chosen

experiment and the individual recordings.

A Scenario describes the conditions of all

experiments proceeded during a course: overground

or treadmill, speed levels, used sensors, number of

trials.

A Proband relates to a real person participating

in one or more experiments. Any proband has an

individual ID "Pxxx".

An experiment relates to one scenario and one

proband. It groups all trials and recordings of the

chosen experiment. Each experiment is identifiable

by a unique number.

A Test/Trial is related to a walk of the proband.

Each test is characterized by a unique number.

Each Sensor has a unique key in the form of

[F/S/T/VP] [L/R] or EMG or XYZ, where F - Foot, S

- Shank, T - Thigh, VP - Pelvis, L - Left, R - Right,

EMG - sEMG, XYZ – accelerometer, e.g. FsL or SR.

http://gaitanalysis. th-brandenburg.de

3.2 Webpage

A webpage

serves the entry point - the graphical

user interface - to the GaitAnalysisDataBase. The

webpage offers two informative subpages, a

download area and three approaches to search data

sets via scenarios, probands or experiments.

Tab “Scenarios”: all scenarios extended by a

table of related experiments are listed.

Tab “Probands”: all anonymized subjects, their

personal data and a table of scenarios, where they

were involved, is presented.

Tab “Experiments”: the table of all experiments

provides links to individual pages where all datasets

sorted by the number of trials and the sensors are

listed. Any entry includes the abbreviation of the

sensor position, stride characteristics and the number

of strides as well as links to the preview and the

download of the selected dataset.

3.3 Recordings

Each record consists of two parts: the metadata block,

describing the data, and the binary coded data.

3.3.1 Xsens MTw Recordings

Xsens IMUs record acceleration, angular velocity and

magnetic rate vectors, as well as the orientation of the

sensor. The acquired data were transformed from the

sensor related coordinate system into the world

coordinate system and gravity compensated.

Each dataset contains the measurements from a

single IMU. The metadata block includes: date,

proband ID, number of test, sensor position, sensor id,

number of columns, number of samples, precision,

sampling rate, duration of recording, number of speed

levels, scaling factors of acceleration, angular

velocity and quaternions.

3.3.2 PLUX Recordings

PLUX units sample up to 8 channels depending on

the number and the type of primary sensors. While a

sEMG has only one signal, an XYZ sensor has three

channels - one for each direction.

Each EMG or XYZ dataset contains the

measurements of a single unit. The metadata block

includes: date, alias, number of test, type of sensor,

number of columns, abbreviations of muscles,

number of samples, sampling rate, precision, duration

of recording, number of speed levels.

BIOSIGNALS 2020 - 13th International Conference on Bio-inspired Systems and Signal Processing

168

3.4 Export and Import of Data Files

Any recording can be downloaded from the specified

experiment page.

The names of the data files are transliterated as

YYYMMDD_Pnnnn_ttt_XXX_mmm,

where YYYYMMDD - date, Pnnnn - proband ID,

ttt - trail, XXX - sensor type and , mmm - sensor id,

e.g. . 20141029_P0008_001_FsL_125.dat.

All data sets could be downloaded at once if

meaningful.

To import the downloaded data files into the

MATLAB workspace, two functions are provided.

4 EXEMPLARY APPLICATION

The datasets included in the database were acquired

in study and research projects for different objectives.

First of all, we – students and lecturers - have learned

a lot about walking and gait, about sensors and signal

processing, about good practice when conducting

experiments with individuals, documenting them,

collecting and storing data, and, last but not least,

making the data available for public use.

Anyway each series of experiments was

originated and motivated by specific questions. Two

of them will be presented in the following sections.

One typical testing in gait analysis is the

“Catwalk” scenario where the subjects are asked to

pass several times a distance at a constant, self-

selected pace. The distance depends on the available

equipment and space. In standard clinical conditions

the subject walks about 10 m in one direction, so that

at least 4 normal strides could be observed. If mobile,

wearable sensors are used, the distance could be

extended to 20 m (indoor) or more (outdoor). In this

case the number of observable gait cycles is greater

than 10 and an averaging could make sense.

In various studies the subjects are asked to walk

“normal, slow and fast”, with “self-selected

comfortable, reduced or increased velocity”, i.e. on

three discriminable speed levels what the subject can

easily perform, replicate and control. It should be

mentioned that there are large differences between the

“normal” speeds of persons, because it depends on

physiological characteristics, e.g. height, weight, age

and gender, and for one subject in dependence of the

personal situation (daily routine, walk with friends,

shopping.)

4.1 Individual Pace Control

Examples of investigated questions are:

 What velocities are chosen by any individual as

comfortable, reduced and increased?

 Is the individual able to control their walking

pace, to separate normal, slow and fast, and to

replicate them?

 How do the individuals realize the increase of the

pace – by increasing the stride length or the

cadence or both simultaneously?

Figures 4a and b include two types of images.

Above the determined average stride velocities,

length and cadence are shown for both the left and the

right shank sensors and for each of the trials. Below

their frequency distribution is represented in form of

histograms added by the mean value (standard

deviation).

Figure 4a: Individual stride velocity, length and cadence

(above) and their histogram (below) of P0100.

Figure 4b: Individual stride velocity, length and cadence

(above) and their histogram (below) of P0107.

Figure 4 presents two different cases:

 In figure 4a the three levels are clearly separated.

The change of pace is a result of varying both the

length and the cadence was changed.

 Figure 4b presents the case where the normal and

slow are on the same level.

A Public Dataset of Overground and Treadmill Walking in Healthy Individuals Captured by Wearable IMU and sEMG Sensors

169

4.2 Overall Velocity Distribution

In 2014 and 2015 two groups were involved in the

scenario B70Z12, B57E12 and B77E12 “Repetitive

Indoor Walking”. The participants were asked to pass

four times a distance of about 20 meters at constant

pace, first at a self-selected comfortable velocity, then

with reduced and last with increased speed. 18

subjects were collected for the test group -

experiments 001-018, October/November 2014, age

28 (12), height 178 (9). 10 persons created the control

group – experiments 052, 053-062, October 2015, age

24 (9), height 177 (11).

Figure 5: Velocity histogram of the test group (above), the

control group (central) and both groups (below).

Seven Xsens MTw sensors were applied: one in the

middle of the back, 6 pairwise above the knee, above

the ankle and on the shoe. The only difference

between the experiments in 2014 and 2015 concerns

to the position of the foot sensors: lateral centric in

2014 and on the metatarsus in 2015.

Figure 5 shows the velocity distribution of the test

(above) and control group (below) and all together. It

could be mentioned that the results (mean value and

the standard deviation) are very close. The small

difference could be a result of the different positions

of the foot sensors. The distance between the foot

sensor positioned lateral and the hip is obviously

greater than the correspondent distance from the

metatarsus (on the top of the foot). It follows that the

measured accelerations and the calculated velocities

are a bit higher in 2014.

4.3 Dependencies of Gait Parameters

on Walking Velocity

There are various ways to investigate the dependency

of stride characteristics on the walking velocity. The

scenarios of “Repetitive Walking” or “Walking on a

treadmill with pre-set speeds or incremental speed

profile” could be employed. Here the scenario

L70Z01 – “Treadmill with incremental speed profile”

was chosen. 7 speed levels run through, including at

least 40 seconds periods of stabilized walking, during

each trial. The disadvantage of this approach could be

the influence of the moving belt on the walking

pattern.

Figure 6: Influence of the treadmill speed on stride

characteristics: length, height, width, velocity, strike and

lift angle, duration of stride, stance and swing (red – left,

blue – right leg).

Figure 6 illustrates the dependency of stride

characteristics like stride length, height, width and

velocity, strike and lift angle, duration of stride,

stance and swing on the numbers of steps for a single

experiment (Loose, 2016). The number of executed

strides corresponds to the belt velocity which was

incremented every 60 seconds by 1 km/h and

BIOSIGNALS 2020 - 13th International Conference on Bio-inspired Systems and Signal Processing

170

decremented later every 30 seconds. It increases

together with treadmill speed (see subplot “stride

velocity”). Obviously stride length, height, strike and

lift angle rise with increasing stride velocity, while

stride, stance and swing duration descend. A higher

stride velocity is achieved by increasing the stride

length and shortening the stride duration. The

relationship between the stance and swing phases is

changing. The stance phase becomes shorter

relatively to the swing phase.

5 CONCLUSIONS

This paper presents the GaitAnalysisDataBase, which

provides data sets of walking for public use, that

could be used to develop and evaluate algorithms, and

to investigate different research problems without

having to collect own data.

The paper gives a detailed description of the

database, its webpage and the used terms (scenario,

proband, experiment and trial). An overview about

the experimental setup, the acquisition of data, the

procedure of experiments, data processing and

evaluation is included. Results of exemplary

investigations are described in the second part of the

paper.

The public database opens up manifold

opportunities for research and development tasks as

well as for educational projects and studies in the field

of gait analysis.

To extend the database including so-called

annotation files, providing intermediate results, i.e.

detected gait events, duration and length of strides, is

still under consideration. The presented results on any

experiment page are not intended for comparison, but

for classification of the trial.

We would be pleased to see the offer taken up and

would appreciate any hint to improve or extend the

database. We kindly ask you to cite this paper should

you use this database for your publications or the

research.

ACKNOWLEDGEMENTS

We would like to thank all volunteers, students,

employees and colleagues who participated in or

contributed to our gait studies in various roles as

proband, experimenter, evaluator, software developer

or supervisor.

REFERENCES

Bovi, G., Rabuffetti, M., Mazzoleni, P., Ferrarin,M. (2011).

A multiple-task gait analysis approach: Kinematic,

kinetic and EMG reference data for healthy young and

adult subjects. Gait & Posture 33 (2011) 6-13.

Ferber R, Osis ST, Hicks JL, Delp SL. 2016. Gait

biomechanics in the era of data science. Journal of

Biomechanics 49:3759_3761 DOI 10.1016/j.jbiomech.

2016.10.033.

Fukuchi, C.A., Fukuchi, R.K., Duarte, M., 2018. A public

dataset of overground and treadmill walking kinematics

and kinetics in healthy individuals. PeerJ 6:e4640; DOI

10.7717/peerj.4640

Goldberger, A.L., Amaral, L.A.N., Glass, L, Hausdorff,

J.M., Ivanov, P.Ch., Mark, R.G., Mietus, J.E., Moody,

G.B., Peng, C-K, Stanley, H.E., 2000. PhysioBank,

PhysioToolkit, and PhysioNet: Components of a New

Research Resource for Complex Physiologic Signals.

Circulation 101(23): e215-e220 [Circulation Electronic

Pages; http://circ.ahajournals.org/content/101/23/e215.

full]; 2000 (June 13)

Hnat SK, Moore JK, Van den Bogert AJ. 2015. Command

treadmill motions for perturbation experiments.

Available at https://zenodo.org/ record/ 16064#.WtU4

U62ZPBI.

Kirtley C. 2014. CGA normative gait database. Available at

http:// www.clinicalgaitanalysis.com/ data/ .

Knudson D. 2017. Confidence crisis of results in

biomechanics research. Sports Biomechanics

16:425_433 DOI 10.1080/14763141.2016.1246603.

Loose, H., (2015). Gait Patterns in Standard Scenarios:

Using Xsens MTw Inertial Measurement Units.

Proceedings of REM 2015, Bochum, Germany, 18-20

November 2015.

Loose, H., Orlowski, K., Tetzlaff, L. (2016): Estimation of

the Average Gait Velocity based on Statistical Stride

Parameters of Foot Sensor Data. In Proceedings of the

9th International Joint Conference on Biomedical

Engineering Systems and Technologies, ISBN 978-

989-758-170-0, pages 277-283. DOI: 10.5220/00058

22602770283

Perry, J. (1992). Gait Analysis - Normal and Pathological

Function. Slack Inc.

Roetenberg, D., Luinge, H., and Slycke, P. (04/2009).

Xsens mvn: Full 6 dof human motion tracking using

miniature inertial sensors. XSENS TECHNOLOGY,

White Paper.

Willson JD, Kernozek T. 2014. Gait data collected at

university of wiscosin-la crosse. Available at https://

innsport.com/ related-products/ data-sets/uw-l-gait-

data-set.aspx.

Winter D.A., Knowledge base for diagnostic gait

assessments. Medical Progress Through Technology

19:61_81.

Winter D.A., Biomechanics and Motor Control of Human

Movement, WILEY, 2009.

A Public Dataset of Overground and Treadmill Walking in Healthy Individuals Captured by Wearable IMU and sEMG Sensors

171