Provided by the author(s) and University College Dublin Library in accordance with publisher
policies. Please cite the published version when available.
Title Using a foot mounted accelerometer to detect changes in gait patterns
Authors(s) Patterson, Matthew; Caulfield, Brian
Publication date 2013-07-07
Conference details 2013 35th Annual International Conference of the IEEE Engineering in Medicine and
Biology Society (EMBC), Osaka, Japan, 3 - 7 July 2013
Publisher IEEE
Item record/more information http://hdl.handle.net/10197/7449
Publisher's statement © © 2013 IEEE. Personal use of this material is permitted. Permission from IEEE must be
obtained for all other uses, in any current or future media, including reprinting/republishing
this material for advertising or promotional purposes, creating new collective works, for
resale or redistribution to servers or lists, or reuse of any copyrighted component of this
work in other works.
Publisher's version (DOI) 10.1109/EMBC.2013.6611286
Downloaded 2022-08-10T00:36:14Z
The UCD community has made this article openly available. Please share how this access
benefits you. Your story matters! (@ucd_oa)
© Some rights reserved. For more information, please see the item record link above.
Seediscussions,stats,andauthorprofilesforthispublicationat:http://www.researchgate.net/publication/257602504
Usingafootmountedaccelerometertodetect
changesingaitpatterns.
ARTICLEinCONFERENCEPROCEEDINGS:...ANNUALINTERNATIONALCONFERENCEOFTHEIEEEENGINEERINGIN
MEDICINEANDBIOLOGYSOCIETY.IEEEENGINEERINGINMEDICINEANDBIOLOGYSOCIETY.CONFERENCE·JULY2013
DOI:10.1109/EMBC.2013.6611286·Source:PubMed
DOWNLOADS
37
VIEWS
48
2AUTHORS:
MatthewRPatterson
ShimmerResearch
17PUBLICATIONS30CITATIONS
SEEPROFILE
BrianCaulfield
UniversityCollegeDublin
167PUBLICATIONS1,359CITATIONS
SEEPROFILE
Availablefrom:MatthewRPatterson
Retrievedon:17June2015
Abstract— The purpose of this study is to investigate how data
from a foot mounted accelerometer can be used to detect motor
pattern healthy subjects performed walking trails under two
different conditions; normal and stiff ankle walking. Lower
body kinematic data were collected as well as accelerometer
data from both feet. An algorithm is presented which quantifies
relevant swing phase characteristics from the foot
accelerometer. Peak total acceleration during initial swing was
significantly higher in the stiff ankle condition (M = 33.10, SD =
5.12) than in the normal walking condition (M = 29.47, SD =
5.75; t(7) = 4.32, p = .003, two-tailed). There was a large effect
size (eta squared = 0.853). Time between peak acceleration
during initial swing to foot strike was significantly shorter in
the stiff ankle condition (M = 0.42, SD = 0.02) than in the
normal condition (M = 0.44, SD = 0.03; t(7) = -2.54, p = .039,
two- tailed). There was a large effect size (eta squared = 0.693).
Simple to process metrics from tri-axial accelerometer data on
the foot show potential to detect changes in ankle kinematic
patterns.
I. INTRODUCTION
raditional gait analysis tools are expensive and take a
significant amount of time to obtain data from [10].
These factors severely limit how often gait analyses can
be performed on a patient and also limit the number of
patients gait analysis can be performed on. The complexity
and cost associated with traditional gait analysis techniques
it is important for research to address issues concerning the
use of wearable sensor technology which may allow gait
analysis to be accessible to more patients in easier to use and
deployable applications outside the laboratory [7].
A significant body of work in the ambulatory monitoring
field has gone into using inertial measurement units to
determine kinematic data during various movements [11,
12]. This approach provides useful data in an easier to use
set-up than traditional measurement techniques. However,
the fact that sensors are required on each body segment
limits such an approach from being applied to deployable,
every-day monitoring applications. Long-term patient
monitoring has traditionally obtained metrics such as activity
recognition, calorie counting or step-counting. While these
metrics are very useful for many clinicians, there is an
*This work was supported by Science Foundation Ireland under grant
07/CE/I1147.
M. Patterson is with the CLARITY Centre for Sensor Web Technologies
and the School of Public Health, Physiotherapy and Population Science,
University College Dublin, Belfield, Dublin 4, Ireland. Phone +353 1
7166500, email: matt.patterson@ucd.ie.
B. Caulfield is with the CLARITY Centre for Sensor Web Technologies,
the TRIL Centre, and the School of Public Health, Physiotherapy and
Population Science, University College Dublin, Belfield, Dublin 4, Ireland
(email: b.caulfield@ucd.ie).
opportunity to obtain more detailed quality of movement
data from the sensors that are used to obtain these gross
metrics.
Using shoe embedded sensors is a promising opportunity
for ambulatory gait monitoring, since it does not require a
user to apply any extra sensors. Foot mounted sensors have
been used most commonly to assess spatio-temporal
parameters of gait [18]. Accelerometers at the foot have
been used to detect foot strike (FS) [6]. The addition of a
gyroscope allows for toe-off (TO) identification as well as
measurement of step length [5, 8, 9]. Spatio-temporal
parameters are useful for identifying important phases of the
gait cycle, for calorie estimation and for identifying large
changes in movement patterns. However, kinematic changes
in gait are not always associated with changes in step time
and step length. Early disease or injury development could
potentially be detected by identifying changes in lower body
kinematics that are not necessarily associated with spatio-
temporal changes in gait.
Previous work has shown the potential for shoe mounted
accelerometers to predict lower body kinematic patterns
during gait in a long-term monitoring scenario [13]. This
approach only worked well when a complex calibration was
performed on each patient. Perhaps, replicating traditional
measurement tools is not necessary. Perhaps, the inertial
data from the foot on its own could be used to determine if
an abnormal gait pattern exists.
Capturing gait metrics with inexpensive sensors in an
unobtrusive manner, while people go about their daily lives
would allow for more patients to have their gait analyzed on
a regular basis [1]. This may allow more people to live
healthier lives and decrease health-care costs by allowing for
earlier intervention in disease and injury development [21].
The aim of this investigation is to determine if data from a
foot mounted accelerometer can be used to identify change
in a person’s gait patterns.
II. METHODS
Eight participants were recruited for the study; six female
and two male. Ethical approval was granted by the
Universities ethical review board and each subject signed an
informed consent form. The participants average age was
27.4 years (+/- 2.67 years), their average weight was 59.1
kgs (+/- 12.4 kgs) and their average height was 1.68m (+/-
0.11m).
Each subject performed ten 15m walking trials in a
biomechanics laboratory under two conditions; normal
walking and a simulated stiff ankle gait. Stiff ankle gait was
simulated by use of a lace up ankle brace which restricted
ankle plantar-flexion. Both conditions were done at the same
Using a foot mounted accelerometer to detect changes in gait
patterns
Matthew R. Patterson, Student Member IEEE and Brian Caulfield, Member, IEEE
T
walking speed. Walking speed was determined at the end of
each trial by finding stride distance and stride time for the
right heel marker and dividing them together. An average
walking speed was determined from the normal walking
trials and stiff ankle trials were only included if they were
within 0.20 m/s of the normal walking average. The stiff
ankle condition was included to attempt to replicate diseases
that result in limited ankle range of motion; such as
Parkinson’s disease, stroke, diabetes mellitus and cerebral
palsy [14, 15]. Ankle sprains and injuries can also result in
abnormal plantar-flexion activity during walking. Subjects
wore their normal, everyday shoes during the walking trials.
A CODA motion capture system (Charnwood Dynamics,
Leicestershire, UK) was used to collect kinematic data.
Markers were placed on the participants right and left sides
at the following locations, PSIS, ASIS, greater trochanter,
femoral condyle, fibular head, lateral malleolus, heel and 5
th
metatarsal. An IMU (Xsens MTx, Enschede, Netherlands)
was placed on the dorsal aspect of each subjects shoe above
the shoe laces, held in place with athletic tape. The sensor
was placed on the dorsum on each foot so that the distal
aspect of the sensor lined up with a perpendicular line
coming from the 5
th
metatarsal (Fig 1).
Figure 1. Inertial sensor placement on the dorsum of the foot. The local x
and y acceleration were used to find FS as well as to quantify aspects of the
swing phase during gait.
A. Accelerometer processing
Accelerometer data was analyzed using MATLAB 2009b
(Mathworks, Massachusetts, USA). Gyroscope and angle
data were not used because the purpose of this study was to
see if accelerometer data alone could be used to determine
changes in gait patterns. Total acceleration (TA) was
calculated from x, y and z acceleration signals by using
equation 1.
Total acceleration (TA) = sqrt (Ax
2
+Ay
2
+Az
2
) (1)
The 3-axis data were transformed due to the fact that
looking at specific axes is very sensitive to how the sensor is
mounted on the shoe. Very small changes in mounting
location could be erroneously flagged as a change in
movement if the data were to be looked at on each axis
individually. This is a trade off and it has been chosen to
pick usability at the expense of accuracy. TA cannot show
what is happening in which axis, but by using TA it means
that any user can mount the sensor on their shoe properly
and will be able to generate useful results. This is important
for the case where a user is mounting a sensor on their shoe
or a manufacturer is creating the shoe with a 3-axis
accelerometer built in.
An algorithm was created to quantify aspects of the
acceleration signal from each walking trial post test. First,
the fundamental frequency of TA was determined and then
TA was band pass filtered between 0.3 Hz to the
fundamental frequency. On the sinusoidal resultant curve,
positive going zero crossings were used to estimate where
initial swings were and negative going zero crossings were
used to estimate where FS occurred. Initial swing peaks
were found in a range around the positive going zero
crossing point and FS was found according to a previously
published method [6].
Two variables were calculated from the accelerometer
data. Peak TA during initial swing (PTAIS) and the time
between PTAIS to FS (TTAFS). Figure 1 shows a typical
TA curve during a gait cycle and where these features occur.
PTAIS is of interest because it shows how quickly the foot is
being moved just after TO. Abnormal movement patterns
around TO are likely to result in different acceleration
patterns during initial swing and PTAIS may be able to pick
these up. TTAFS is a timing variable which likely has a
scalar relationship to swing time, but requires less
processing and sensors to find since it does not include TO.
Figure 2. TA data from the right foot for a typical swing cycle and the
quantified accelerometer variables. The dark limbs on the stick figure
represent the right arm and leg.
B. Statistics
The accelerometer variables were averaged over each of
the five trials for each condition for each subject. A two-
tailed, paired samples t-test was used to test the null
hypothesis that the accelerometer variables could not detect
the change in gait between the constrained gait conditions
and their speed matched controlled conditions.
A two-tailed, paired samples t-test was also used to
investigate differences in walking speed and ankle
kinematics to see if there was a significant difference in
these factors between the conditions. Associated effect sizes
(eta squared) were calculated and quantified according to
Field as 0.10 = small effect size, .030 = medium effect size
and 0.50 = large effect size [3]. The level of significance was
set at p < 0.05.
III. RESULTS
There was not a statistically significant difference in
walking speed between the normal (M = 1.422, SD = .141)
and stiff ankle walking condition (M = 1.366, SD = .195),
t(7) = 1.55, p = .165 (Table 1).
Peak ankle plantar-flexion was significantly decreased the
normal walking condition (M = -23.96, SD = 3.33)
compared to the stiff ankle walking condition (M = -15.05,
SD = 3.34), t(7) = -5.477, p < 0.05 (Table 2).
Figure 3. Peak ankle plantar flexion during initial swing compared
between the normal and stiff ankle trials for each subject. Ankle movement
was restricted in all subjects in the stiff ankle condition.
PTAIS values were significantly higher in the stiff ankle
condition (M = 33.10, SD = 5.12) compared to the normal
walking condition (M = 29.47, SD = 5.75; t(7) = 4.32, p =
0.003, two-tailed, Figure 4). There was a large effect size (eta
squared = 0.853). TTAFS values were significantly lower in
the stiff ankle condition (M = 0.42, SD = 0.02) compared to
the normal walking condition (M = 0.44, SD = 0.03; t(7) = -
2.54, p = 0.039, two-tailed, Figure 5). There was a large
effect size (eta squared = 0.693). Walking speed and hip and
knee kinematics were not altered between the conditions.
Ankle kinematics, PTAIS and TTAFS were changed between
the conditions (Tables 1 & 2).
Figure 4. Change in PTAIS for each subject from the normal walking
condition to the stiff ankle walking condition. Error bars show standard
deviations.
Figure 5. Change in PTAIS for each subject from the normal walking
condition to the stiff ankle walking condition. Error bars show standard
deviations.
TABLE I. TOTAL ACCELERATION AND SPATIO-TEMPORAL
GAIT METRICS
PTAIS*
TTAFS*
Speed
Stride
length
Stride
time
Units
m s-1 s-1
sec
m s-1
m
sec
Stiff
ankle
29.47
(5.74)
.436 (.028)
1.35
(.204)
1.493
(157.6)
1.12
(.103)
Normal
walking
33.01
(5.12)
.421 (.022)
1.43
(.141)
1.535
(118.2)
1.08
(.068)
Mean values per group (SD), ** indicates significant difference between
groups
TABLE II. KINEMATIC GAIT METRICS
Peak
ankle
plantar-
flexion *
Peak ankle
dorsi-
flexion
stance
Peak ankle
plantar-
flexion ang
velocity*
Peak ankle
dorsi-flexion
angular
velocity*
units
deg
deg
deg s-1
deg s-1
Stiff
ankle
-15.03
(3.37)
10.21 (4.76)
-25.63 (5.87)
14.48 (2.96)
Normal
walking
-23.96
(3.33)
11.54 (4.17)
-35.07 (6.85)
20.15 (3.76)
Mean values per group (SD), ** indicates significant difference between
groups
17
22
27
32
37
42
1 2 3 4 5 6 7 8
PTAIS (m/s/s)
Subject
normal stiff ankle
0.36
0.38
0.4
0.42
0.44
0.46
0.48
1 2 3 4 5 6 7 8
TTAFS (sec)
Subject
normal stiff ankle
