=Paper=
{{Paper
|id=Vol-2815/CERC2020_paper19
|storemode=property
|title=Wearable Gait Analysis – Stepping Towards the Mainstream
|pdfUrl=https://ceur-ws.org/Vol-2815/CERC2020_paper19.pdf
|volume=Vol-2815
|authors=Philip Morrow,Huiru Zheng,Graham McCalmont,Haiying Wang,Sally McClean
|dblpUrl=https://dblp.org/rec/conf/cerc/MorrowZM0M20
}}
==Wearable Gait Analysis – Stepping Towards the Mainstream==
https://ceur-ws.org/Vol-2815/CERC2020_paper19.pdf
COVID-19 Research and Smart Healthcare
Wearable Gait Analysis – stepping towards the
mainstream
Philip Morrow, Huiru Zheng, Graham McCalmont, Haiying Wang, and Sally
McClean
School of Computing, Ulster University,
Jordanstown Campus,Shore Road,
Newtownabbey,Co. Antrim, BT37 0QB, Northern Ireland,
{morrow-p4, h.zheng, mccalmont-g, hy.wang, si.mcclean}@ulster.ac.uk
Abstract. Wearable technologies have transformed the accessibility of gait
analysis, offering the opportunity to venture outside of the laboratory and into
everyday life. This research article is concerned with investigating the progress
that has been made, and the steps that remain in gait analysis becoming a main-
stream activity. The evidence for the effectiveness of wearable gait analysis
technologies was reviewed, indicating that these devices are capable of support-
ing gait analysis in a ‘real-world’ environment. Research into the application of
wearable technology for gait analysis was found to limited in terms of scope,
with progress still to be made in improving the perception of these devices.
Challenges to be addressed within this field of research were identified: (1)
Large scale data collection; (2) Broader scope of wearable gait analysis; (3)
‘Real-world’ gait analysis; (4) Case study research approach (5) Gait analysis as
a service; (6) User testing/evaluation.
The development of wearable gait analysis systems, and the underlying re-
search that supports their application, should be cognizant of how mainstream
acceptance is contingent upon meeting these challenges. The path towards ad-
dressing them is considered in the context of the eZiGait portable gait analysis
system, highlighting the value of collaboration with industry.
Keywords: Gait analysis, wearable devices, inertial measurement unit, smart insoles, clin-
ical application, gait rehabilitation, industry perception.
1 Introduction
Gait Analysis is concerned with the study of human motion in order to develop an
understanding of an individual’s ability to walk. A normal gait pattern has a positive
effect on quality of life by enabling a person to perform their everyday activities un-
inhibited [1]. This ability is assessed by analysis of specific characteristics that con-
stitute their walk pattern. The analysis of the gait pattern of an individual involves the
extraction of specific parameters, such as spatiotemporal parameters, that are associ-
Copyright © 2020 for this paper by its authors.
CERC 2020 306 Use permitted under Creative Commons License
Attribution 4.0 International (CC BY 4.0).
COVID-19 Research and Smart Healthcare
2
ated with different characteristics of walking. The different phases of walking con-
stitute an overall gait cycle that represents the activity between one foot contacting the
ground and that same foot again contacting the ground. Depending on the level of
granularity used, the gait cycle can be composed of up to eight phases [2]. At the
coarsest level of granularity, the two main phases are the stance phase, where the foot
remains in contact with the ground, and the swing phase where the foot is lifted off
the ground. Identifying different phases of the gait cycle enables spatiotemporal pa-
rameters to be extracted e.g. walking speed, step time, step length etc. The use of
pressure and force sensors enables kinetic parameters such as centre of mass and cen-
tre of pressure to be determined [3].
The development of gait analysis methods has largely been concerned with
healthcare and rehabilitation. The role of analysing the gait of an individual in
providing insight into their general health has long been established [4]. Conditions
that have been determined to result in gait disorders include Cerebral Palsy [5], Par-
kinson’s Disease [6] and Alzheimer’s Disease [7]. For rehabilitation, gait analysis
can contribute to improving an individual’s limb movements to enable them to regain
a normal walking function. The applications of gait analysis within sports include
improving performance [8] and avoiding injury [9]. For security purposes, individual
gait profiles have been demonstrated to provide the basis for identifying a person
[10]. The range of applications continues to expand, with artificial gait for humanoid
robots and more realistic computer-generated models for the entertainment industry
among recent developments [11].
Traditionally gait analysis took the form of human observation of an individual
subject walking. This observation was performed by a trained professional and could
be complemented with self-reporting by the subject. General quantitative features
such as walking speed, as well severe gait disorders, can be discerned through this
approach. These judgements are inherently subjective in nature and thus limit their
reliability in informing a medical diagnosis or treatment decision. More significantly,
subtle deviations from ‘normal’ gait cannot be discerned from human observation
limiting the insight that can be provided by this analysis. More sophisticated gait
analysis approaches were developed to facilitate more objective analysis of gait. For
example, motion laboratories incorporating such techniques as video imaging to rec-
ord movements enable analysis of kinematic information. This generally involves the
placement of markers on the body of the test subject, with marker-less systems yet to
gain as widespread use [12]. The use of single or multiple cameras determines wheth-
er 2D or 3D motion analysis can be performed [13]. Force plates measure vertical
ground forces exerted by an individual to enable analysis of kinetic information [14].
The primary downside of such approaches was the considerable cost incurred in set-
ting up an appropriate laboratory to conduct the analysis. The prohibitive cost in turn
ensured that access to this form of gait analysis was limited to specialist locations.
The advent of wearable technology greatly reduced the barriers to use present with
conventional technology, thus opening the possibility of widening access to gait anal-
ysis. Despite the breakthrough provided by these wearable devices mainstream ac-
ceptance of gait analysis remains a goal yet to be achieved. This research article con-
siders the developments made within the field of wearable gait analysis and the barri-
307 CERC 2020
COVID-19 Research and Smart Healthcare
3
ers that remain in fully realizing its potential. Potential approaches that may be
adopted to move towards mainstream acceptance are then discussed in the context of
the eZiGait mobile gait analysis system that is currently being developed.
2 Review of Wearable Gait Analysis Research/Development
The review of developments within this area commences with a comparative over-
view of the main types of gait analysis technologies that have been developed thus
far. The evidence for their effectiveness of wearable devices for enabling gait analy-
sis is then considered. The contribution that these devices can make is then examined
in terms of the scope of previous research and the potential users of such technology.
2.1 Comparative Overview of Gait Analysis Technologies
Table 1 provides an overview of the main types of technologies that are used to
perform gait analysis. Optical Motion Capture systems represent the most accurate
technology for capturing Kinematic information, but correspondingly incur the high-
est costs in terms of initial setup and subsequent operation. Force plates systems, that
measure Vertical Ground Forces, can be implemented at varying levels of cost. Both
systems are limited to use within a laboratory environment and only cater for record-
ing data over limited durations. The application of gait analysis is therefore limited to
controlled exercises that may not be reflective of the ‘normal’ walking patterns exhib-
ited by an individual person. Gait Pressure mats have been developed as a more prac-
tical alternative, with medium costs incurred and some degree of portability. The
relative bulk of such systems still inhibits their accessibility to some extent, and they
remain suitable for indoor use only. The limited length of the mats means that they
are unsuitable for extended recording durations. The development of wearable tech-
nologies represents the most significant breakthrough in terms of accessibility. The
main types of inertial sensors provide measurements in up to 3 axes of motion and can
be incorporated together in the form of an Inertial Measure Unit (IMU). Accelerome-
ters provide measurements of the rate of change of velocity for an individual in their
own real frame. Gyroscopes provide measurements of orientation and angular veloci-
ty. Magnetometers measure the direction and strength or relative change of a magnet-
ic field at a location. These inertial sensors may be utilised within a standalone de-
vice or within other technology such as a smartphone. The presence of inertial sen-
sors within smartphones provides the greatest degree of availability, although
standalone IMU devices provide greater flexibility e.g. can be worn on different parts
of the body. The location of the IMU device impacts both the measurement accuracy
achievable and the level of comfort for the user. The costs of such devices can vary
according to how they are implemented, but a significant reduction can be achieved
relative to lab-based systems. The suitability for outdoor use and capacity for captur-
CERC 2020 308
COVID-19 Research and Smart Healthcare
4
ing sensor data over an extended duration provides a significant advantage in terms of
measuring ‘real-world’ data.
Table 1. Comparison of Main Technology Types
Technology Type Key Characteristics Practicality
Optical Motion Capture - High Measurement Precision - Lab-based System
System - High Technology Costs - Requires expert
- High Computational Costs operation
- Limited recording
duration
Force Plates - Low Measurement Precision - Limited Portability (due
- Low to High Technology Costs to weight) and used
- Low Computational Costs mainly within Lab
environment
- Limited recording
duration
Gait/Pressure Mat - Variable Measurement - Portable but suitable for
Precision indoor use only
- Medium Technology Costs - Requires expert
- Low Computational Costs operation
- Limited recording
duration
Inertial Measurement - Low Measurement Precision - Portable use both
Unit (Accelerometer, - Low to Medium Technology indoors and outdoors
Gyroscope, Costs - Can be worn and
Magnetometer) - Low Computational Costs operated by user
- Level of comfort
dependent upon design
- Extended recording
duration
Insole Pressure Sensor - Low Measurement Precision - Portable use both
System - Low to Medium Technology indoors and outdoors
Costs - Can be worn and
- Low Computational Costs operated by user
- Extended recording
duration
Insole Pressure Sensor systems provide a more accessible alternative to Force
Plates and at a significantly lower cost. These insoles deploy an array of sensors to
record pressure at several locations on the sole of each foot. The number of sensors
embedded into an insole can range from a few to over 40 depending on the design.
309 CERC 2020
COVID-19 Research and Smart Healthcare
5
The level of availability of Insole Pressure Sensor systems is not as widespread as that
of smartphone-based inertial sensors, but they otherwise retain the same benefits in
terms of measuring ‘real-world’ data. Smart insole systems may combine the use of
inertial sensors with pressure sensors, enabling a wider variety of application for any
recorded data as well as the potential for fusion of the different types of data. In these
systems, the inertial sensors may be attached to the insoles or embedded directly into
the insoles. The development of these smart insole systems has addressed many of
the shortcomings of previous gait analysis systems.
2.2 Clinical Application of Gait Analysis
Jarchi et al. [15] reviewed gait analysis involving accelerometery with a focus on how
it is applied to clinical applications. The review included 159 research papers, start-
ing from the year 2000, and defined three classifications for these studies. The most
common category, at 38% of research papers, was the validation of gait parameters
against more established approaches such as video analysis, force-plates etc. The
other two categories were focused on applying gait analysis to clinical applications,
with 32% of papers utilising an accelerometer and 30% of studies incorporating single
or multiple configurations of sensors.
Baker et al. [16] described four main reasons for performing clinical gait analysis.
The first reason was the diagnosis of a specific disease, followed by the assessment of
the severity of that disease or an injury. The other two reasons were concerned with
either monitoring or predicting the progress of the disease or injury, with or without
an intervention taking place on the patient. The success of clinical gait analysis can
be judged according to the extent that it supports the desired outcomes for these in-
tended uses i.e. its efficacy.
Table 2 shows how research involving the clinical efficacy of gait analysis has
progressed over the past two decades. This research was focused on conventional
three-dimensional instrumented gait analysis performed using laboratory methods
such as motion capture and force plates. Two systematic reviews, conducted by the
same authors [17] - [18], categorised the research studies published during their re-
spective preceding decade according to the efficacy type that was addressed. For the
second systematic review, a new efficacy type (2b) was introduced to reflect how the
focus of research had evolved. This type 2b research was concerned with an evalua-
tion of the efficacy of treatment at a group level, in contrast to individual patient out-
comes addressed at type 5. In some cases, the research conducted in an individual
study incorporated multiple adjacent efficacy types. The overall number of research
studies had risen significantly in the latter decade, from 210 to 2712 papers, indicating
the increased significance of this field of research. The level of efficacy that has been
investigated most has also changed from type 1 to type 2, including the new type 2b.
This may reflect the progress beyond evaluating the technical performance of gait
analysis methods towards their clinical application. The efficacy type 2b research
indicated the significance of being able to diagnose patients as being part of a group
CERC 2020 310
COVID-19 Research and Smart Healthcare
6
and consequently predicting treatment outcomes for that type of patient. The absence
of any increase in research studies at higher efficacy types indicates the limitations on
the understanding of how gait analysis can contribute at these levels. In particular,
the question to be asked is whether the limitations on accessibility associated with
laboratory gait analysis have themselves impacted the progress of research in this
regard.
Table 2. Number of Research Studies Grouped by Efficacy Type
Efficacy Type Number of Research Number of Research
Studies (Jan 2000 to Sept Studies (Sept 2009 to Oct
2009) 2019)
1 - Technical 116 313
2 - Diagnostic Accuracy 89 1466
2b - Outcome Prediction - 927
3/4 - Diagnostic Thinking 11 6
& Treatment
5 - Patient Outcome 7 3
6- Societal 1 0
The highest level of efficacy (type 6) was concerned with the cost effectiveness of
gait analysis in terms of its impact on society. The need for such research was high-
lighted in a conference held amongst professionals within this field to establish a con-
sensus on the general progress of gait analysis [19]. The absence of cost effectiveness
studies on the use of motion laboratories was broadly agreed to be a significant prob-
lem. The absence of standardised procedures for gait analysis and publicly available
data on ‘normal’ gait profiles was also identified as a problem to be addressed. The
scientific evidence supporting the use of gait analysis was generally considered lim-
ited to a few medical conditions. The overriding conclusion of the conference was
that the value of gait analysis as a research tool has thus far exceeded its value for
clinical applications.
2.3 Overall Effectiveness of Wearable Gait Analysis Devices
Numerous research studies have examined the performance of wearable devices in the
context of Gait Analysis. These studies have sought to demonstrate the feasibility of
utilising wearable sensors for gait analysis and provide a means of evaluating them as
a substitute for more conventional laboratory approaches.
Kobsar et al. [20] conducted a systematic review of studies on the validity and reli-
ability of wearable inertial sensors. The context of this review was on healthy adults
walking as opposed to individuals with any underlying medical conditions. The re-
view covered research papers from 1998 to 2019 in terms of set parameters, rating
each parameter according the quantity, quality and consistency of results across stud-
ies that addressed the parameter. For the step time and stride time parameters the
results were rated as excellent. For step length, stride length, swing time and stance
311 CERC 2020
COVID-19 Research and Smart Healthcare
7
time the results were rated as good to excellent. For gait variability and gait sym-
metry parameters the results were only rated as poor to moderate. The issue with the
lower performing parameters was more an issue of the limited number of studies, or
their design, rather than the performance of the sensors themselves. The main weak-
nesses of the research studies were from a statistical perspective e.g. underpowered
results, unjustified sample sizes, and inadequate statistical analysis in support of the
evaluation of results. The meta-analysis undertaken within this review highlighted
the difficulty in comparing results across studies. The large number of studies were
completed without a standardised protocol and addressed different subsets of gait
parameters. For each gait parameter there were typically between three to five studies
that could be compared. Establishing a comprehensive appraisal of the validity and
reliability of wearable inertial sensors was concluded to more a question of quality
than of quantity in terms of future research studies.
Direct evidence for the performance of inertial sensor-based systems relative to
that of conventional gait analysis approaches was reviewed for the time period from
2005 to 2017 [21]. In this systematic review, a total of 16 research articles were se-
lected that compared gait parameters obtained using wearable or inertial sensors
against those obtained using a motion laboratory. Meta-analysis was performed on
seven different gait parameters addressed by multiple studies. Table 3 shows a sum-
mary of the results of this meta-analysis.
Table 3. Meta-Analysis of Gait Parameter Results
Gait Num- Number Standard- 95% Confidence
Parameter ber of of Subjects ised Mean Interval (Leftmost
Studies Difference value, Rightmost
value)
Gait Speed 5 149 0.11 (-0.10, 0.33)
Step Length 6 245 0.14 (-0.14, 0.43)
Step Time 5 234 -0.03 (-0.24, 0.18)
Stance Time 3 140 0.50 (0.24, 0.76)
Stride Time 3 110 -0.03 (-0.29, 0.23)
Cadence 4 138 0.46 (-0.41, 1.34)
Swing Time 3 140 -0.63 (-1.32, 0.05)
Standardised Mean Difference (SMD) refers to the difference in the mean value of
each group (IMU and motion laboratory) divided by the standard deviation. In this
case, a positive SMD indicates that the IMU group had the greater mean value and a
negative SMD indicates that the other group had the greater mean value. The leftmost
and rightmost values of the 95% confidence interval for the SMD value provides an
indication of the level of uncertainty associated with the value. Statistically signifi-
cant differences between the groups are indicated when the 95% confidence interval
does not contain zero. The results indicate that for gait speed, step length, step time
and stride time there was a good level of agreement between the two groups. Only
stance time demonstrated a statistically significant difference between the two groups.
CERC 2020 312
COVID-19 Research and Smart Healthcare
8
Cadence and Swing Time had the greatest level of uncertainty regarding the SMD
value, so caution is necessary in drawing any conclusions for those parameters.
The value of being able to obtain the gait parameters using relatively low-cost
wearable devices in short walking exercises was noted by the authors of the review.
Limitations on the understanding of the effectiveness of wearable inertial sensors that
can be derived from this review were, however, identified. These included the differ-
ent devices, algorithms and test procedures used by the studies. The use of wearable
devices was not considered to be a substitute for motion laboratory gait analysis as the
latter approach is necessary for identifying the locomotor strategy used by patients.
The use of Artificial Intelligence (AI) approaches such as machine learning for
analysis of gait data has been a common research theme. The effectiveness of AI
approaches to assist with gait analysis methods was considered by a systematic re-
view of research studies focused on inertial sensors and adaptive algorithms [22].
The review, considering only Journal Papers, looked at research studies published
between 1968 and 2016 that used IMUs and adaptive AI algorithms to classify gait
events. The quality of results obtained from adaptive AI algorithms was reported to
be above average, leading the authors to conclude that they are suitable for use in gait
analysis. The lack of standardization in terms of the methods used was again high-
lighted as an issue moving forward. As most of the studies utilised healthy test sub-
jects, the potential for improvements with patients whose gait is impacted by a medi-
cal condition was not discernible from the research.
The use of inertial sensors for gait recognition was reviewed by Sprager and Juric
[23]. The performance of inertial sensor-based gait recognition was found to be more
effective for large datasets, in the region of 200 people, with error rates of around 5%
reported. These research studies generally used very short walking exercises, typical-
ly involving only a few steps, thus limiting the extent to which they would be repli-
cated in practice. For those research studies that allowed for walking in an uncon-
trolled environment over a longer period the reported accuracy rates ranged from
under 70% to above 90%.
2.4 Scope of Wearable Gait Analysis Devices
Chen et al. [24] conducted a systematic review of gait analysis research studies in-
volving wearable sensors, examining how pervasive the technology could become.
The authors asserted that only 0.2% of gait analysis papers both involved the use of
inertial sensors and were conducted in real clinical settings. There was no indication
that the total number of research papers accounted for those published before inertial
sensors were developed so this relative proportion may be understated. The review
included 2906 papers focused on applying gait analysis to medical conditions. The
distribution of these studies was found to strongly favour some medical conditions
with Parkinson’s Disease (29%), Cerebral Palsy (17%), Orthoses (13%), Lower Limb
Osteoarthritis (6%) and Post Stroke (6%) accounting for over 70% of the research
papers. The remaining approx. 30% of research was distributed across 13 categories
suggesting that research gaps remain to be addressed. Possible reasons suggested for
this pattern of research were limited availability of affordable technology or a need to
313 CERC 2020
COVID-19 Research and Smart Healthcare
9
align technology advancements with the medical knowledge within those domains. A
systematic review of the use of smartphone systems for physical rehabilitation re-
vealed a similar narrow focus [25]. From the 74 research studies that were reviewed,
the diseases that dominated the research were stroke, cardiac disease, balance im-
pairment and joint/limb rehabilitation.
In order to provide the greater accessibility within standardised medicine that
wearable sensors are intended to provide it is necessary to gain acceptance within the
medical community. In terms of research, using established gait analysis procedures
ensures generalisability for the results. Numerous research studies have shown that
standard clinical tests such as the Timed Up and Go (TUG) test and the 6-Minute
Walk test, can be successfully administered using low cost sensors in mobile or wear-
able devices [26], [27]. The Standardised procedure outlined for the TUG test has
also been successfully adapted towards use with wearable devices [28]. The usage
of wearable devices for standard tests of a longer duration, as with the 6-Minute Walk
test, has also enabled a more detailed analysis of gait features to be performed. For
example, sensor data obtained from turning during the test provides a clearer indica-
tion of impaired gait than may be obtained from regular walking in a straight line.
Research into the use of wearable devices for gait analysis outside of the laboratory
environment was reviewed by Benson et al. [29]. The research papers were com-
prised of 43 walking studies, 13 running studies and a single study incorporating both
forms of exercise. While the outcomes of the studies were assessed as being suffi-
ciently reported they generally did not adequately describe the statistical power of
their results. There was some progress towards using larger number of participants
within walking studies, but it was recommended further improving this by monitoring
gait over longer periods of time and in natural environments. The need to address the
usability of wearable sensors was also highlighted. For long term use, the location the
device is to be worn and its size/weight need to be as unobtrusive as possible without
compromising the validity of the measurements that can be obtained.
3 Adoption of Wearable Gait Analysis Devices
Investigating the perception of wearable devices provides insight into the progress
that has been made towards achieving mainstream acceptance. The development of
wearable technology has been largely driven by potential medical applications, but in
terms of impact the main market has been the general fitness industry [30]. The med-
ical field represents a higher barrier to entry in terms of performance in comparison to
the general fitness market. The level of accuracy acceptable for a consumer-oriented
device would not generally be considered sufficient for supporting medical decisions.
In contrast, familiarity with gait analysis is much less prevalent within general con-
sumers.
Issues affecting the adoption of wearable gait analysis tools are discussed in the fol-
lowing sections. This considers the perception of wearable devices amongst both
professional users, interested in adopting them into their work practices, and individu-
CERC 2020 314
COVID-19 Research and Smart Healthcare
10
als that would have their gait characteristics analysed. An example of how gait analy-
sis can be integrated into the wider healthcare system is then provided.
3.1 Perception of Wearable Devices
In a survey of the adoption of wearable sensors in the workplace [31], 90.4% of re-
spondents that wore a device at work did so for monitoring general physical activity.
The most popular devices that respondents expressed confidence in, such as “Fitbit”,
were typically providing easy to understand data such as step counts. Doubts were
commonly expressed by respondents about the validity and efficacy of more sophisti-
cated uses of wearable technology.
Mobile phone health apps have experienced substantial growth, but a systematic
review of the scientific evidence behind their diagnostic performance was under-
whelming [32]. Even when including research focused on symptom monitoring, for
supporting diagnosis, it was found that studies were lacking both in terms of quantity
and quality.
The need for a patient-centred focus has been promoted as a response to the in-
creasingly complex delivery of healthcare. In the context of physical rehabilitation,
several themes have been identified within research [33]. It is important that treat-
ment is individualized for the patient and that they have a good understanding of their
symptoms and treatment path. Goal setting and establishing a feeling of empower-
ment for the patient were reported as having a positive effect in helping them cope
with their condition. The use of wearable devices offers the potential for providing
patients with this patient-centred treatment both in and outside of physiotherapy ses-
sions.
Morris et al [34] conducted a survey of clinicians concerning their perspectives on
the use of mobile health and rehabilitation applications. Over 500 clinicians were
surveyed, drawn from a range of professions such as physical, occupational and psy-
chological therapists. While a large proportion of the respondents reported prescrib-
ing specific out of clinic exercises and interventions to patients, it was clear that ac-
ceptance of mobile solutions continues to be a challenge. Only 51% of respondents
expressed comfort in integrating mobile technology into their clinical practice. In
addition, only 23% of respondents considered themselves to be knowledgeable about
the available technology. The top barrier to their use, identified by 72% of respond-
ents, was the inability of patients to learn and correctly use mobile technology.
The perspectives of physiotherapists on the use of wearable or mobile health tech-
nology were investigated by Blumenthal et al. [35] In this study a simplified version
of the popular Technology Acceptance Model (TAM) framework [36] was used to
investigate the willingness of 76 participating therapists and students to implement
these types of technology into their clinical practice. The primary motivation of the
research was to investigate why the usage of this technology remained low within the
physiotherapy industry. The study found no evidence that early adopter behaviour
was influenced by age or previous experience with technology. It was suggested that
the perceived usefulness of the technology was an important determinant of early
adoption. Encouraging physiotherapists to invest their time and resources in imple-
315 CERC 2020
COVID-19 Research and Smart Healthcare
11
menting this technology requires a clear demonstration of how it would add value to
their practice. Increasing patient engagement and improving how progress is com-
municated were rated as highly important by the study participants. The importance
of the user experience was therefore proposed as an important design consideration
for mobile health technologies.
Guillen-Gamez and Fernández [37] investigated the perceptions of the subjects in-
volved in wearable technology research. The study included a total of 606 patients
and relatives, comprised of 60.2% female and 39.8% male participants. The attitude
of the participants towards the usefulness of wearable devices was reported as being
medium high, with a slightly higher level reported for males. In particular, the partic-
ipants viewed the devices as being more useful for caring for the health of elderly
people than for themselves. While men were generally more accepting of the use of
wearable devices there was a greater contrast across age for female participants.
Women under 30 owned more wearable devices than any other group, but women
over 45 had the lowest level of acceptance for the use of such devices. In terms of the
acceptance expressed for the location of the wearable device, on the wrist had the
highest level (Male 52.7%, Female 50.19%) out of five options. The option of weara-
ble devices placed in shoes, relevant for Smart Insole technology, was ranked third for
Male (48.77%) and fourth for Female (45.30 %) participants.
Evaluating usability and accessibility of mobile heath apps has been identified as
an important consideration for achieving a patient-centred focus in improving rehabil-
itation outcomes [38]. In this review, usability was reported to be high for those apps
that were educative and supported self-reporting of symptoms. For apps focusing on
intervention, the most positive effects were found with functional outcomes such as
gait and self-management skills. Positive effects were also found with health out-
comes such as pain and quality of life. Evaluating the impact of mobile health sys-
tems requires consideration of both types of outcomes. Investigating functional out-
comes from using these systems outside of the controlled laboratory environments is
therefore as important as investigating their accuracy.
3.2 Integration of gait analysis into clinical practice
The acceptance of gait analysis within mainstream clinical practice is contingent upon
successful integration into the wider healthcare system. Research on how to achieve
this integration has been limited, but an in-depth study on the design of a gait test
within clinical rehabilitation provided instructive guidelines [39]. This study adopted
a service-oriented approach to design in contrast to viewing gait analysis as a techno-
logical challenge. The multiple stakeholders in this service include patients (and their
relatives), doctors, and therapists. Each of these stakeholders present different needs
that must be met and therefore they must be considered in the design process. The
design in this study was comprised of three main phases. Firstly, ‘User-product prox-
imity effect’ where the gait test is performed in a motion laboratory and the effects on
the various stakeholders are observed. Secondly, ‘Effect and value in the service’
where an overview is obtained of the path followed by the patient through the service
CERC 2020 316
COVID-19 Research and Smart Healthcare
12
e.g. diagnosis, care and treatment decisions. Thirdly, ‘User interactions’ that define
the information flows between stakeholders that are necessary to facilitate implemen-
tation of the gait test. The design recognised the need to treat the patient as an indi-
vidual that will respond in their own way to treatment. Providing guidance to the
patient at each stage of the service being provided is essential in enabling them to
view the gait test as a motivational tool within their overall treatment.
4 Towards Mainstream Adoption of Wearable Gait Analysis
The review conducted in the preceding sections enabled the identification of the
challenges to be addressed for wearable devices. The eZiGait system is introduced as
an example of a wearable gait analysis system, with consideration then given to how
future development of this system can be directed towards addressing the challenges
that have been identified.
4.1 Future Challenges for the adoption of Wearable Gait Analysis Devices
The issues to be addressed on the path towards mainstream acceptance of Wearable
Devices for gait analysis can be summarised as follows:
1. Large scale data collection: in order to achieve a greater understanding of normal
and abnormal gait profiles it is necessary significantly increase the size of datasets.
The accessibility and cost of wearable sensor systems needs to be considered if
they are to enable the establishment of datasets on a large scale. The increase in
the adoption of these systems could potentially decentralise the establishment of
these datasets from researchers to the users.
2. Broader scope of wearable gait analysis: to facilitate the acceptance of gait anal-
ysis it is necessary to broaden the nature of the data collected. Instead of focusing
on a limited selection of pathological conditions research should explore whether
an individual’s gait profile is affected by other conditions. In terms of the sports
and leisure industries it should be investigated whether gait analysis can benefit the
more ‘casual’ athlete rather than being restricted to professional athletes. The im-
pact of wearable gait analysis should be to move gait analysis beyond being re-
stricted to relatively niche markets.
3. ‘Real-world’ gait analysis: the development of wearable sensor devices has ena-
bled the collection of subject data in ‘real-world’ conditions. It is imperative for
research to increase efforts to engage in the collection of data in these conditions.
Wearable sensor technology provides the basis for users to record data in their own
time and location.
4. Case study research approach: increasing the quality of research includes the
adoption of in-depth research studies. Case studies involving the progress of pa-
tients through the treatment/rehabilitation process are necessary to understand both
how gait analysis can inform decisions made at each stage and how it impacts re-
covery for a patient.
317 CERC 2020
COVID-19 Research and Smart Healthcare
13
5. Gait analysis as a service: the value of wearable sensor systems is to be derived
from the service they provide to the various user groups. The objective is therefore
to identify the needs of each user group, such as patients, medical professionals,
athletes etc., and incorporate them into the services provided by the system.
Achieving this objective requires including each user group as stakeholders and
having them guide the direction of development of these services.
6. User testing/evaluation: to ensure the suitability of wearable sensor systems to
fulfil their intended role, they must be subjected to appropriate evaluation by the
intended users so that their design can be refined accordingly. For the wearable de-
vices, it is necessary to ensure that the size/weight and location of the technology
remains comfortable for the patient. For the supporting software, the gait analysis
reports produced must be both be understandable to the users and convey infor-
mation that is of value to them.
The path towards addressing these challenges would include collaboration with or-
ganisations with a vested interest in the growth of wearable gait analysis, such as
medical and rehabilitation organisations. The patients/customers of such organisations
represent prospective users that need to be reached in order to achieve the goal of
wearable technology i.e. to improve their quality of life.
4.2 eZiGait System Overview
The architecture of the eZiGait system is shown in Fig 1. The main components of
the system are the Smart Insoles, the AI Gait Assistant mobile application, and the
Cloud-based Analytics. The Smart Insoles used in previous research studies [40] -
[41] were provided by TreeHouse Technology Ltd in China. There are eight separate
pressure sensors built into each insole, providing readings of Vertical Ground Forces
for eight locations within each foot. A pressure map is thus obtained for each foot,
charting the distribution of weight exerted across each foot during the Gait cycle. An
attached electronic device, that is worn around the ankle, is equipped with an IMU.
This IMU provides a 3-axis accelerometer, 3-axis gyroscope and 3-axis magnetome-
ter. By simultaneously sampling data in three axes, the test subject’s movements are
effectively monitored.
CERC 2020 318
COVID-19 Research and Smart Healthcare
14
Fig. 1. Overview of eZiGait System
The electronic device is equipped with a rechargeable battery that allows sensor
data to be recorded over prolonged exercise sessions. The battery for this device is
charged using any standard USB connection. The sensor readings are streamed in
real-time and transmitted via Bluetooth to any paired device, such as a smartphone.
An API is provided for the Smart Insoles that facilitates communication with other
applications. A mobile app, AIGaitAssistant, has been developed for Android based
smartphones. This app serves as the intermediary between the Smart Insoles and the
Cloud-based Analytics, analysing the gait characteristics of the test subject. The fol-
lowing main functionality is included within the smartphone app:
• Connect to Bluetooth paired Smart Insoles to receive real-time sensor data on test
subject movements.
• Display of real-time sensor data from each of the sensors within the GUI. This
enables the user to verify that each of the sensors is functioning correctly prior to
performing any exercise.
• Record sensor data for a specified exercise session. Each session is defined by the
user in terms of a label to identify that session, the type of exercise to be complet-
ed, automatic or manual upload and generation of a gait report. The recorded ses-
sion data is automatically saved locally on the smartphone as a CSV file.
• Upload the sensor data to a web server via a Wi-Fi connection for subsequent
cloud-based analysis.
319 CERC 2020
COVID-19 Research and Smart Healthcare
15
• Generation of a Gait Report for the completed exercise session. This report is
comprised of the following detail
─ Overall summary of the gait session e.g. time, step count etc.
─ Gait Phase Distribution for each foot, with the overall gait cycle divided into
Loading, Foot-Flat, Pushing and Swing Phases
─ Temporal Symmetry between each foot for Step Time, Stance Time, Swing
Time and Stride Time
─ Pressure Symmetry between each foot for Heel Region, Mid-foot Region, Toe
Region and Overall Foot.
The Cloud-based Analytics are hosted on a web portal that enables further exami-
nation of the recorded sessions by healthcare/rehabilitation professionals. For exam-
ple, the progress of an individual in terms of their gait characteristics can be tracked
over the course of their uploaded sessions. The analysis provided by this web portal
continues to be developed with the aim of applying advanced machine learning ap-
proaches to establish a greater understanding of an individual gait.
4.3 eZiGait and the Future of Wearable Gait Analysis
The several challenges to be addressed for wearable gait analysis to gain more main-
stream acceptance are being considered during the development of eZiGait.
eZiGait is being developed in conjunction with related research projects concerned
with gait analysis. It has been utilised within previous research studies [40] - [41] in
addition to internal data collection and validation. This provides a mechanism for
conducting user testing and refining both functional and non-functional user require-
ments. It also presents opportunities for collaborating with external parties in ‘real-
world’ environments. For example, the Smart Insoles and smartphone app have been
utilised during rehabilitation sessions with a local physiotherapist. This provided the
opportunity to learn from prospective users of the system and refine the user experi-
ence based on the feedback obtained. This ongoing collaboration will provide the
opportunity to collect data from both healthy subjects and rehabilitation subjects. The
healthy subjects include individuals participating in regular sports exercises within a
gym environment. This enables data related to ‘normal’ gait to be collected in a ‘real-
world’ situation in contrast to a lab-controlled environment with specific exercises.
The access to rehabilitation subjects presents a learning opportunity that addresses
multiple challenges.
• The physiotherapist is one the few specialists available in Ireland that provides the
use of an exoskeleton as part of the rehabilitation process. The range of conditions
affecting patients undergoing rehabilitation therefore presents the opportunity to
collect data outside of the common conditions more regularly addressed within re-
search. This will enable an understanding to be developed of how these conditions
impact the gait profile of the subjects.
• The collection of data at successive stages of the rehabilitation process will enable
the subsequent effect of the treatments on the subject’s gait profile to be assessed.
CERC 2020 320
COVID-19 Research and Smart Healthcare
16
The in-depth case-study approach presents the opportunity to learn how gait analy-
sis can form part of the rehabilitation process.
• The input from the physiotherapist throughout the research can guide the develop-
ment of the gait analysis system by communicating which information is of use to
the rehabilitation process and identifying potential new aspects of gait to be inves-
tigated.
• The needs of both the physiotherapist and the rehabilitation subjects, in terms of
the service that is being provided, can be analysed and subsequently specified.
This includes an understanding of the expectations held by these users prior to each
gait analysis session and the level of satisfaction achieved upon completion of the
session. Designing the service is as important to its success as the performance of
the technology.
• The deployment of the system in a ‘real-world’ environment enables user testing
and evaluation of the system to be performed. This will enable the system to be re-
fined in order to enhance the accessibility of software. The presentation of infor-
mation is integral to user understanding and subsequently them benefitting from
their experience with the system.
• The benefits and costs of using the system can be investigated. The impact upon
the rehabilitation process in terms of the extent of the recovery and the timeframe
involved can establish what benefits, if any, that gait analysis provides. The cost
incurred, such as the provision of the technology and the learning time required,
can also be established, enabling an assessment to be made on the value that is
provided.
5 Conclusion
Wearable gait analysis has provided a significant breakthrough in terms of the acces-
sibility and affordability of such technology. This technology provides the potential
for enabling gait analysis to become commonplace within mainstream healthcare
systems and both the sports and leisure industries. Realising this potential is depend-
ent upon directing the technology towards the needs of these users and demonstrating
clear and cost-effective benefits. Research within this field should include investigat-
ing wearable gait analysis systems within the context of those industries that it aspires
to benefit in order to further understand the role that this technology can fill. Bring-
ing wearable gait analysis research out of the lab environment should enable the tech-
nology to follow suit and step forward into the mainstream.
6 Acknowledgement
This research is supported by the Invest NI Proof of Concept (PoC) programme
(Project ID: PcC809).
321 CERC 2020
COVID-19 Research and Smart Healthcare
17
7 References
1. Nutt, J.G., Marsden, C.D., Thompson, P.D.: Human walking and higher-level gait disor-
ders, particularly in the elderly. Neurology. 43(2), 268-279. doi:10.1212/wnl.43.2.268,
(1993)
2. Liu, T., Inoue, Y., Shibata, K.: Development of a wearable sensor system for quantitative
gait analysis. Measurement. 42, 978–988. doi: 10.1016/j.measurement.2009.02.002. (2009)
3. Muro-de-la-Herran, A, Garcia-Zapirain, B, Mendez-Zorrilla, A.: Gait analysis methods: an
overview of wearable and non-wearable systems, highlighting clinical applications. Sen-
sors (Basel). 14(2), 3362-3394. Published 2014 Feb 19. doi:10.3390/s140203362 (2014)
4. Fish, D.J., Nielsen, J.P., Clinical assessment of human gait, Journal of Prosthetics and Or-
thotics 5(2), 39–48, (1993)
5. Tugui, R.D., Antonescu, D.: Cerebral palsy gait, clinical importance. Maedica (Buchar).
8(4), 388-393 (2013)
6. Shichkina, Y, Stanevich, E, Irishina, Y.: Assessment of the Status of Patients with Parkin-
son's Disease Using Neural Networks and Mobile Phone Sensors. Diagnostics (Basel).
10(4), 214. Published 2020 Apr 12. doi:10.3390/diagnostics10040214 (2020)
7. Higuma, M, Sanjo, N, Mitoma, H, Yoneyama, M, Yokota, T.: Whole-Day Gait Monitoring
in Patients with Alzheimer's Disease: A Relationship between Attention and Gait Cycle.
Journal of Alzheimer’s Disease Rep. 1(1) 1-8. doi:10.3233/ADR-170001 (2017)
8. Wahab, Y & Abu Bakar, N.: Gait analysis measurement for sport application based on ul-
trasonic system. Proceedings of the International Symposium on Consumer Electronics,
ISCE. 10.1109/ISCE.2011.5973775. (2011)
9. Silva, B., Natarajan, A., Motani, M. & Chua, K.: A Real-Time Exercise Feedback Utility
with Body Sensor Networks. 49 - 52. 10.1109/ISSMDBS.2008.4575013. (2008)
10. Chen, J.: Gait correlation analysis based human identification. Scientific World Journal.
168275.. doi:10.1155/2014/168275 (2014)
11. Prakash, C., Kumar, R. & Mittal, N.: Recent developments in human gait research: param-
eters, approaches, applications, machine learning techniques, datasets and challenges. Arti-
ficial Intelligence Review. 49, 1-40. 10.1007/s10462-016-9514-6. (2018).
12. Colyer, S.L, Evans, M, Cosker, D.P., Salo, A.I.T.: A Review of the Evolution of Vision-
Based Motion Analysis and the Integration of Advanced Computer Vision Methods To-
wards Developing a Markerless System. Sports Medicine Open. 4(1), 24.
doi:10.1186/s40798-018-0139-y (2018)
13. Klette, R., Tee, G.: Understanding Human Motion: A Historic Review. In: Rosenhahn, B.,
Klette, R., Metaxas, D. (eds) Human Motion. Computational Imaging and Vision, vol 36.
Springer, Dordrech doi:10.1007/978-1-4020-6693-1_1 (2008)
14. Sutherland, D.H.: The evolution of clinical gait analysis part III--kinetics and energy as-
sessment. Gait Posture. 21(4), 447-461. doi: 10.1016/j.gaitpost.2004.07.008 (2005)
15. Jarchi, D., Pope, J., Lee, T.K.M., Tamjidi, L.,Mirzaei, A. and Sanei, S.: "A Review on Ac-
celerometry-Based Gait Analysis and Emerging Clinical Applications," in IEEE Reviews
in Biomedical Engineering, 11, 177-194, doi: 10.1109/RBME.2018.2807182. (2018)
16. Baker, R., Esquenazi, A., Benedetti, M.G., Desloovere, K.: Gait analysis: clinical facts.
European Journal of Physical and Rehabilitation Medicine 52(4), 560-574. (2016)
17. Wren, T.A., Gorton, G.E. 3rd, Ounpuu, S., Tucker, C.A.: Efficacy of clinical gait analysis:
A systematic review. Gait Posture. 34(2), 149-153. doi: 10.1016/j.gaitpost.2011.03.027
(2011)
CERC 2020 322
COVID-19 Research and Smart Healthcare
18
18. Wren, T., Tucker, C., Rethlefsen, S., Gorton, G. & Ounpuu, S.: Clinical efficacy of in-
strumented gait analysis: Systematic review 2020 update. Gait & Posture. 80.
10.1016/j.gaitpost.2020.05.031. (2020)
19. Benedetti, M.G., Beghi, E., De Tanti, A., et al.: SIAMOC position paper on gait analysis in
clinical practice: General requirements, methods and appropriateness. Results of an Italian
consensus conference. Gait Posture. 258, 252-260. doi: 10.1016/j.gaitpost.2017.08.003
(2017)
20. Kobsar, D., Charlton, J.M., Tse, C.T.F., et al.: Validity and reliability of wearable inertial
sensors in healthy adult walking: a systematic review and meta-analysis. Journal of Neuro
Engineering Rehabilitation.; 17(1), 62. doi:10.1186/s12984-020-00685-3 (2020)
21. Petraglia, F., Scarcella, L., Pedrazzi, G., Brancato, L., Puers, R., Costantino, C.: Inertial
sensors versus standard systems in gait analysis: a systematic review and meta-analysis.
European Journal of Physical and Rehabilitation Medicine, 55(2), 265-280.
doi:10.23736/S1973-9087.18.05306-6 (2019)
22. Caldas, R., Mundt, M., Potthast, W., Buarque de Lima Neto, F., Markert, B.: A systematic
review of gait analysis methods based on inertial sensors and adaptive algorithms. Gait
Posture. 57, 204-210. doi: 10.1016/j.gaitpost.2017.06.019 (2017)
23. Sprager, S., Juric, M.B.: Inertial Sensor-Based Gait Recognition: A Review. Sensors (Ba-
sel). 15(9), 22089-22127. Published 2015 Sep 2. doi:10.3390/s150922089 (2015)
24. Chen, S., Lach, J., Lo, B. and Yang, G.: Toward Pervasive Gait Analysis With Wearable
Sensors: A Systematic Review, IEEE Journal of Biomedical and Health Informatics, vol.
20(6), 1521-1537, doi: 10.1109/JBHI.2016.2608720. (2016)
25. Moral-Munoz, J.A., Zhang, W., Cobo, M.J., Herrera-Viedma, E., Kaber, D.B.:
Smartphone-based systems for physical rehabilitation applications: A systematic review,
Assistive Technology. 1-14. doi:10.1080/10400435.2019.1611676 (2019)
26. Ponciano, V., Pires, I., Ribeiro, F., Marques, G., Garcia, N., Pombo, N., Spinsante, S., &
Zdravevski, E.: Is The Timed-Up and Go Test Feasible in Mobile Devices? A Systematic
Review. Electronics (Basel). 9(3). 10.3390/electronics9030528. (2020)
27. Storm, F.A., Cesareo, A., Reni, G., Biffi, E.: Wearable Inertial Sensors to Assess Gait dur-
ing the 6-Minute Walk Test: A Systematic Review. Sensors (Basel). 20(9), 2660.
doi:10.3390/s20092660 (2020)
28. Salarian, A., Horak, F.B, Zampieri, C., Carlson-Kuhta, P., Nuttand J.G., Aminian, K.:
iTUG, a sensitive and reliable measure of mobility, IEEE Transactions on Neural Systems
and Rehabilitation Engineering, vol. 18(3), 303–310, (2010)
29. Benson, L.C., Clermont, C.A., Bošnjak, E., Ferber, R.: The use of wearable devices for
walking and running gait analysis outside of the lab: A systematic review. Gait Posture.
63, 124-138. doi: 10.1016/j.gaitpost.2018.04.047 (2018)
30. Seneviratne, S., Hu, Y., Nguyen, T., Lan, G., Khalifa, S., Thilakarathna, K., Hassan, M. &
Seneviratne, A.: A Survey of Wearable Devices and Challenges. IEEE Communications
Surveys & Tutorials. 1-1. 10.1109/COMST.2017.2731979. (2017)
31. Schall, M.C. Jr., Sesek, R.F., Cavuoto, L.A.: Barriers to the Adoption of Wearable Sensors
in the Workplace: A Survey of Occupational Safety and Health Professionals. Hum Fac-
tors. 60(3), 351-362. doi:10.1177/0018720817753907 (2018)
32. Buechi, R., Faes, L., Bachmann, L.M, et al.: Evidence assessing the diagnostic perfor-
mance of medical smartphone apps: a systematic review and exploratory meta-analysis.
BMJ Open. 7(12), e018280. doi:10.1136/bmjopen-2017-018280 (2017)
33. Wijma, A.J., Bletterman, A.N., Clark, J.R, et al.: Patient-centeredness in physiotherapy:
What does it entail? A systematic review of qualitative studies. Physiotherapy Theory and
Practice. 33(11), 825-840. doi:10.1080/09593985.2017.1357151 (2017)
323 CERC 2020
COVID-19 Research and Smart Healthcare
19
34. Morris, J., Jones, M., Thompson, N., Wallace, T., DeRuyter, F.: Clinician Perspectives on
mRehab Interventions and Technologies for People with Disabilities in the United States:
A National Survey. International Journal of Environmental Research and Public Health.
16(21), 4220. doi:10.3390/ijerph16214220 (2019)
35. Blumenthal, J., Wilkinson, A., Chignell, M.: Physiotherapists' and Physiotherapy Students'
Perspectives on the Use of Mobile or Wearable Technology in Their Practice. Physiother-
apy Canada. 70(3), 251-261. doi:10.3138/ptc.2016-100.e (2018)
36. Davis, F.D.: Perceived ease of use, and user acceptance of information technology. Man-
agement Information Systems Quarterly. 13(3), 319–40. https://doi.org/10.2307/249008
(1989)
37. Guillen-Gamez, F. & Fernández, M.J.: Empirical Study Based on the Perceptions of Pa-
tients and Relatives about the Acceptance of Wearable Devices to Improve Their Health
and Prevent Possible Diseases. Mobile Information Systems. 10.1155/2019/4731048.
(2019)
38. Nussbaum, R., Kelly, C., Quinby, E., Mac, A., Parmanto, B., Dicianno, B.E.: Systematic
Review of Mobile Health Applications in Rehabilitation. Archives of Physical Medicine
and Rehabilitation. 100(1), 115-127. doi: 10.1016/j.apmr.2018.07.439 (2019)
39. Marín, J., Blanco, T., Marín, J.J., Moreno, A., Martitegui, E., Aragüés, J.C.: Integrating a
gait analysis test in hospital rehabilitation: A service design approach. PLoS One. 14(10),:
e0224409. doi: 10.1371/journal.pone.0224409 (2019)
40. McCalmont, G et al.: eZiGait: Toward an AI Gait Analysis and Assistant System, IEEE
International Conference on Bioinformatics and Biomedicine (BIBM), Madrid, Spain,
2018, pp. 2280-2286, doi: 10.1109/BIBM.2018.8621176. (2018)
41. An, G et al.: Feasibility study of a smart insole based AI Gait Assistant system, BIBE
2019; The Third International Conference on Biological Information and Biomedical En-
gineering, Hangzhou, China, 1-5. (2019)
CERC 2020 324