Professional Documents
Culture Documents
1 Prototype For The Estimation and Evaluation of Walking Velocity Using Acceleration Transducers
1 Prototype For The Estimation and Evaluation of Walking Velocity Using Acceleration Transducers
Abstract—This work proposes the development of an statistic comparative analysis of the used devices and twenty
electronic prototype with a low cost, based on the utilization of simulations.
acceleration micro-electric-mechanical sensors (MEMS) which
make possible to obtain the biomechanical information from II. BIBLIOGRAPHIC REFERENCES
movements when walking and so permit to obtain estimation
about the average velocity. The system which has been put in Every person exerts a movement which follows a standard
practice captures the information from the sensors through the (style) or even a repetitive continuity of such process [1] [2]
micro-controller and sends them through an intuitive interface, [3]. Therefore, can be established the recognition of an
designed in the GUIDE environment of Matlab software. With individual because of the way of walking as well as the
the intention of validating the developed prototype it was personality through an analysis of the rhythm that the person
proposed a new methodology in order to calculate the walking uses when walking.
velocity, however, in this phase are used platforms which were
certified acquired, developed by National Instruments and which Two of the physical magnitudes that can be monitored
have direct connection with the LabVIEW environment. Twenty through the acceleration sensors and spin are the linear
tests were made simulating the walking using an ergometric acceleration and the angular velocity with which is possible
treadmill with two different velocities, obtaining for the first the calculation of velocity, linear movement and the rotation
prototype an estimated average error of 2.66% and 3.19% (1st angles during the walking movement [4] [5].
and 2nd Velocities, respectively) and with the validation prototype
an error of 1.38% and 2.25% from the acquisitions. The accelerometer has a significant less consumption of
energy than gyroscopes and, besides that it is trustworthy
Keywords—MEMS accelerometer, ATMega 2560 Arduino, when considering a measurement with a lower level of
Matlab, NI SCXI 1000 Chassis, PCI 6072 and LabVIEW. variation along time [6]. Bruxel, Zaro, Palhano, Avila and
Strohaecker [2], studied the choosing of the proper parameters
I. INTRODUCTION when obtaining the data to apply to tibialaccelerometry. It is
The proposed work here implements a prototype which use possible to say, according to [2] [7], that the contribution of
two accelerator sensors fixed on the leg in order to obtain an energy of each component of the signal, it is proportional to
estimation of walking velocity of human beings. Since the the square of the spectrum of frequencies (|F(w)|2). The total
system has to work not only with electronic devices, first of energy is the summation of energy in each band of frequency
all, it was necessary to study human cinematic movement as and it is indicated by the total area under the curve. It is called
well as to what is the proper methodology that has to be used as spectral density of the power of signal (DEP or PSD), it
for the acceleration analysis when waking with the pace of a makes possible to evaluate what of frequencies that are on the
regular walking.. signal carry the biggest amount of energy.
In the next chapters will be approached the constructive Figure 1 introduces the proposed model in the study carried
methods for the prototype. Chapter III has been split in four out by Shuozhi Yang and Quingguo Li [8], where were
stages: the first one makes part of a presentation of the characterized twelve different positions in order to relate them
mathematical model applied to the feedback from the sensors; to the position where the accelerometer(s) is (are) installed
the second one refers to the mechanical, electrical (Hardware) with the respective magnitudes to be measured. With the
and intuitive (Software) materials used when assembling the attention focused for the calculation of walking velocity, in
prototype; the third stage is the experimental technology of this model the authors classify the fixation place with the
tests, that is it, it makes possible to compare the implemented acceleration and/or the angular velocity besides of choosing
prototype as well as the validation device, it was necessary to three types of sensors: the one-dimensional (uniaxial), two-
use a walking treadmill and to know, starting from a dimensional (biaxial) and three-dimensional (triaxial). In this
calculation what is the real velocity of the individual who is work the walking rod will be assembled with acrylic material
moving on the equipment; in the fourth stage it was applied an and fixed exactly with its center on the place (S2) indicated in
the model.
According to E. Bishop and Q. Li [4] [16], instead of Considering the angle of leg θ (t),according to [4], can be
estimating indirectly special parameters, one alternative is to mapped including the angular velocity w(t), as it is shown in
determine the movements using the time included in the (5):
measured accelerations. Thus, acceleration measures are !
necessary and to know at each given moment the angle of the = " + 0 (5)
device in relation to the gravity to subtract the gravitational
When the momentum isθ (0), it is possible to accept this
value for the initial angle of the rod before to be included. In
order to calculate the movements along the horizontal line of
Cartesian plane and the coordinates of vertical axes, the signal
of accelerations ax (t) and ay (t) can be reorganized according
to velocitiesVx (t) andVy (t), which correspond to the global
system of coordinates shown in (6) and (7):
# = " + $ 0 (6)
# = " + $ 0 (7)
20
in account the twenty different simulations executed with the
2
Accelerometer Lower
15
Accelerometer Above same individual. This number of samples was empirically
10
determined for the project since E. Bishop´s works and Q. Li´s
5
works [4] have together with other bibliographic references
0
-5
[2] [32] [16],which use similar values from fifteen to eighteen
-10
walking simulations.
-15
Balbinot and Brusamarello [33] assume that data follow a
-20
familiar curve in the area of science and technology being
-25 called curve of normal distribution which makes it possible the
0 100 200 300 400 500 600 700 800 900 1000
verification of frequency of what happens considering a given
5
datum and the measurement of data distortion. To calculate
Normal Acceleration (m/s )
2
Accelerometer Lower
Accelerometer Above
0 the data dispersion measurement the used properties are the
-5 variance σ? and the standard deviationσ
?, shown in (15).
-15
2
# 1 4 = #
52 ∗ 26 × 89: 2<<
/60
(14)
3
The experimental methods used in this work can be split in
two different groups. The first one is related to the real
velocity measured on the treadmill, that is, with Equation 14 it
was measured the real theoretical performance that the
individual has in the two possible different options when
moving on the equipment and so, obtaining 2.45 m/s for the
first value and 3.56 m/s for the second value.
Making part of the second group of the experimental
Figure 5: Matlab Software of Estimated Walking Speed.
model, the reached final velocity of the implemented
prototype as well as the equivalent device for the validation The consequences of the experimental methodology used
stage suggested in the project. Units MEM´S in all tests are for each one of both velocities of the treadmill were obtained
mounted on the constructed rod in the first module of the from the respective final average velocities for each one of
methodology of working being fixed to a distance of 10 cm both prototypes. Considering the experimental measurements
distant from each other while it is fixing on the individual in Table II, the next step was to put in practice a study related
exactly on point S2 as shown highlighted in Figure 1. to the real result that the devices generated.
TABLE II.
ESTIMATION OF WALKING VELOCITY WITH TWO IMPLEMENTED PROTOTYPES
TREADMILL Prototype LIS344ALH/Arduino2560/Matlab Validation of Prototype
Experimental Tests (Estimation of Average Velocity m/s)
1º - 2.29 6º - 2.48 11º - 2.27 16º - 2.29 1º - 2.51 6º - 2.42 11º - 2.41 16º - 2.45
Velocity 1 (2.46 2º - 2.17 7º - 2.15 12º - 2.18 17º - 2.39 2º - 2.42 7º - 2.44 12º - 2.26 17º - 2.36
m/s) 3º - 2.49 8º - 2.44 13º - 2.43 18º - 2.25 3º - 2.33 8º - 2.40 13º - 2.39 18º - 2.40
4º - 2.15 9º - 2.19 14º - 2.36 19º - 2.42 4º - 2.31 9º - 2.44 14º - 2.50 19º - 2.44
5º - 2.21 10º - 2.45 15º - 2.44 20º - 2.26 5º - 2.37 10º - 2.41 15º - 2.41 20º - 2.31
Average 2.32 2.41
Stand. Deviation 2.66 (%) 1.38 (%)
Experimental Tests (Estimation of Average Velocity m/s)
1º - 3.45 6º - 3.49 11º - 3.29 16º - 3.45 1º - 3.48 6º - 3.42 11º - 3.51 16º - 3.55
Velocity 2 2º - 3.15 7º - 3.47 12º - 3.40 17º - 3.40 2º - 3.48 7º - 3.45 12º - 3.32 17º - 3.34
(3.56m/s) 3º - 3.53 8º - 3.46 13º - 3.13 18º - 3.24 3º - 3.42 8º - 3.58 13º - 3.30 18º - 3.33
4º - 3.17 9º - 3.58 14º - 3.48 19º - 3.46 4º - 3.48 9º - 3.32 14º - 3.20 19º - 3.54
5º - 3.27 10º - 3.19 15º - 3.48 20º - 3.56 5º - 3.48 10º - 3.47 15º - 3.55 20º - 3.40
Average 3.387 3.46
Stand. Deviation 3.19 (%) 2.25 (%)
Figure 6 shows the comparative of final average velocity processing in the signal of accelerometers but in the final
between the two prototypes, the one developed with the result, there is a similar behavior in the average as well as in
working methodology (blue frame) as well as the one of the standard deviation.
validation (yellow frame). Starting from zero velocity, the In order to obtain the real average value which was
grey broken line represents the ideal velocity obtained from measured with both devices it was executed the evaluation of
the conveyor belt of treadmill, therefore, it is possible to combined uncertainness included in the whole measurement
obtain the evaluation of two points of reference when the process. Thus, the calculation of error obtained only from the
average velocity of the individual on the treadmill with standard deviation, the uncertainness is included due to the
velocity 1 (2.46 m/s) and velocity 2 (3.56 m/s) for the resolution of the instrument (µr) as well as the generated
comparative of both of them and the estimated ones made by uncertainness due to the variability of tests (µr). The existent
the device. errors due to the inference of the resolution of the analogic
digital converser (A/D) were calculated as shown in (16)
4
where the resolution is an intrinsic component for the least
3.46 significant bit of the acquisition:
3,5
μM = 1Resolution/2/√34 (16)
3
3.38
In the case of an experimental variation it is possible to
Speed Estimate (m/s)
2,5 2.41 obtain the calculation as shown in (17) where N is the number
of simulations which were executed [33]:
2 ?
σ
2.32 μS = / 0 (17)
√J
1,5
V. DISCUSSION AND CONCLUSSION
Treadmill Speed
1 At the end of this document were achieved the main aimed
goals during the different stages of the methodology of the
0,5 proposed project and thus it was able to validate the produced
prototype in order to leave as the legacy of this project a
0 platform with a low cost (close to 96.88% less) of project and
0 1 2
Treadmill Velocity (m/s)
3 4 which also represents the certified devices of validation [26]
[27] ]28], which uses an accelerometer sensor with a 100
Figure6: Comparing Velocities of Prototypes. times higher sensibility than the implemented prototype in this
In the calculation of the difference which was found work [23] [30]. Even when working in real time with the
between two average velocities the value is 0.09 m/s (Velocity validation platform, the results kept constant and this identifies
1) and 0.08 m/s (Velocity 2). Considering the difference of the same result observed in [5]: a no relevant collecting of data
measurements for the standard deviation collected from 20 in real time for this type of acquisition and mathematical
simulations (N = 20), the reached value is, respectively, 1.28% processing of data.
and 0.94%, that is, such a difference in the average can be In works [4] [2] [32] and [16], the errors of measures
originated by the specific validated prototype, a dedicated assumed values from 5% to 8% being that in [6] it was used
equipment (high financial cost) can exerts differentiated one single accelerometer and the calculated error resulted
14%. However, these works don´t take in account two 11. YEOH, W.; PEK, I.; YONG, Y.; CHEN, X.; WALUYO, A.
“Ambulatory monitoring of human posture and walking speed using
fundamental parameters for the final measure: the real
wearable accelerometer sensors”. In Proceedings of 30th Annual
mathematical evaluation of velocity of the treadmill and the InternationalConference of the IEEE Engineering in Medicine and
existent errors when a combined uncertainness of Biology Society, Vancouver, BC, Canada, 20–24 August 2008; pp.
measurement is included in the comparative calculation 5184–5187.
12. AMINIAN, K.; ROBERT, P.; J´EQUIER, E.; SCHUTZ, Y.
between the different velocities.
“Estimation of speed and incline of walking using neural network”.
For the constructed prototype here the µr uncertainness IEEE 1995, 44, 743–746.
assumes 2.88% for the error of measure while µr represents 13. MIYAZAKI, S. “Long-term unrestrained measurement of stride length
0.60%. Adding up the total evaluated uncertainness the final and walking velocity utilizing a piezoelectric gyroscope”. IEEE Trans.
Biomed. Eng. 1997, 44, 753–759.
error for the calculated device for the velocity 1 of the
14. AMINIAN, K.; NAJAFI, B.; BULA, C.; LEYVRAZ, P.; ROBERT, P.
treadmill is 3.96%; in a comparative with the suggested results “Spatial-temporal parameters of gait measured by an ambulatory
by the bibliographic reference [16], there was a significant system using miniature gyroscopes”. J. Biomech. 2002, 35, 689–699.
difference in the result of the final error, something that 15. TONG, K.; GRANAT, M. “A practical gait analysis system using
gyroscopes”. Med. Eng. Phys. 1999, 21, 87–94.
proved the efficiency for the mathematical model which was
16. DONELAN, J. M.; LI, Q.; NAING, V.; HOFFER, J. A; WEBER, D. J.;
used and also for the offered results during the experimental KUO, A. D. “Biomechanical energy harvesting: generating electricity
stages. In a universe of a bigger discussion, the resultant final during walking with minimal user effort”. Science, 319(5864):807–10,
values can be more accurate because as the measure is 2008.
17. SABATINI, A.M.; MARTELLONI, C.; SCAPELLATO, S.;
increased the number of experimental simulations, the error
CAVALLO, F. “Assessment of walking features from foot inertial
per each experiment decreases and consequently the estimated sensing”. IEEE Trans. Biomed. Eng. 2005, 52, 486–94.
velocity is reached [33]. 18. HUANG, C.; LIAO, Z.; ZHAO, L. “Synergism of INS and PDR in self-
Acquisition, processing and management of data from the contained pedestrian tracking with a miniature sensor module. IEEE
Sensors”.2010, 10, 1349–1359.
gait of walking systems require high investment analysis. Thus
19. MARTIN, S.; VAN A.; BATEN, C.; VELTINK, P. “Ambulatory
the prototype developed in this work offers in addition to estimation of foot placement during walking using inertial sensors”. J.
validated information, a platform which can be used anywhere Biotech. 2010, 43, 3138–3143.
in the human body. For example, with this platform is possible 20. OJEDA, L.; BORENSTEIN, J. “Non-GPS navigation for security
personnel and first responders”. J. Navig. 2007, 60, 391–407.
to measure the impact on the individual's knees, and get an
21. BEBEK, O.; SUSTER, M.; RAJGOPAL, S.; FU, M.; HUANG, X.;
analysis of the cervical position and consequences of irregular CAVUSOGLU, M.; YOUNG, D.; MEHREGANY, M.; VAN Bogery,
walking can cause. A.; MASTRANGELO, C. “Personal navigation via high-resolution
gait-corrected inertial measurement units”. IEEE Trans. Instrume.
REFERENCES Meas. 2010, 59, 3018–3027.
22. MARIANI, B.; HOSKOVEC, C.; ROCHAT, S.; BULA, C.;
1. GRIFFIN, M. J. “Handbook of Human Vibration”. Sand Diego:
PENDERS, J.; AMINIAN, K. “3D gait assessment in young and
Academic Press, 1996.
elderly subjects using foot-worn inertial sensors”. J. Biomech. 2010.
2. BRUXEL, Y.; ZARO, M. A.; PALHANO, R.; AVILA, A. O. V. ;
23. STMICROELECTRONICS LIS344ALH 2008. MEMS inertial sensor
±2/±6 ultracompact
STROHAECKER, T. R. . Selection of the proper parameters when
high performance 3-axis linear
acquiring data through the computer: application on
accelerometer..Revision 1. 2008 (datasheet).
TibialAcelerometryTecniouro Magazine, number 245, pages 72-77,
24. ATMEL ARDUINO MEGA 2560. Doc ID 2549P. “8-bit Atmel
July of 2009.
Microcontroller with 64k/128k/256k/ Bytes In-System Programmable
3. WINTER, D. A. “Biomechanics and Motor Control of Human
Flash”. Augusto, 2012.
Movement”. 2 ed. Ontario, Canada: University of Waterloo Press,
25. GAFANIZ, A.; LOPES, G.; PIRES, P. “Cinematic Analysis of
1990.
Walking”. Biomechanical Movement Subject, Undergraduate Course in
4. BISHOP E. e LI Q.“Walking speed estimation using shank-mounted
Biomechanical Engineering, 2005.
accelerometers”. IEEE International Conference on Robotics and
26. NATIONAL INSTRUMENTS. “Chassis NI SCXI-1000, NI SCXI-
Automation, May 3-8,2010, Anchorage, Alaska, USA, 2010.
1000DC, NI SCXI-1001 forModule SCXI – 1530 / SCXI – 1531”.
5. LI, Q.; YOUNG, M.; NAING, V.; DONELAN, M. “Walking Speed
National Instruments, 2008.
and Slope Estimation Using Shank-mounted Inertial Measurement
27. NATIONAL INSTRUMENTS. “Module SCXI – 1530 / SCXI – 1531,
Units”. IEEE 11th International Conference on Rehabilitation Robotics,
Channel Dynamic Signal Acquisition”. National Instruments, 2008.
2009.
28. NATIONAL INSTRUMENTS. “NI - PCI 6221, NI - PCI 6072”.
6. ZIJLSTRA, W.; HOF, A. “Assessment of spatial-temporal gait
National Instruments, 2008.
parameters from trunk accelerations during human walking”. IEEE
29. BRÜEL & KAER. Calibration document and characteristics of the
2003, 18, 1–10.
three axes of the Endevco 65-100 sensors, Serial Number 11979, March
7. LATHI, B. P. “Signal Processing & Linear Systems”. Berkeley
of 2007.
Cambridge Press. 1998.
30. BRÜEL & KAER. Calibration document and characteristics of the
8. YANG, S.; LI, Q. “Inertial Sensor-Based Methods in Walking Speed
three axes of the Endevco 65-100 sensors, Serial Number 11978, March
Estimation: A Systematic Review”. ISSN 1424-8220, 2012.
of 2007.
9. ALVAREZ, J.; GONZALEZ, R.; ALVAREZ, D.; LOPREZ, A.;
31. IMINIPA.Tachometer Picture / MDT Digital Contact 22338A
RODRIGUES, J. “Multisensory approach to walking distance
(Handbook of instructions. Iminipa 15, January of 2003.
estimation with foot inertial sensing”. In Proceedings of 29th Annual
32. BRUXEL, Y.; SCHEPP, L. G.; PALHANO, R.; ZARO, M.;
InternationalConference of the IEEE Engineering in Medicine and
LOUREIRO. “TibialAcelorometry: Processing of Signal”; Tecnicouro,
Biology Society, Lyon, France, 23–26 August 2007; pp. 5719–5722
v. 30, pages 72-77, 2009.
10. SONG, Y.; SHIN, S.; KIM, S.; LEE, D.; LEE, K. “Speed estimation
33. BALBINOT, A.; BRUSAMARELLO V. J. 2011 Instrumentation and
from a tri-axial accelerometer using neural networks”. In Proceedings
fundamentals of measures. Rio de Janeiro: LTC, 2011. Vol. 2. ISBN
of 29th Annual International Conference of the IEEE Engineering in
978-85-216-1879-9.
Medicine and Biology Society, Lyon, France, 23–36 August 2007; pp.
3224–3227.