Information 07 00047

See discussions, stats, and author profiles for this publication at: https://www.researchgate.
net/publication/305399251
An mHealth Tool Suite for Mobility Assessment
Article in Information (Switzerland) · July 2016

DOI: 10.3390/info7030047
CITATIONS READS
11 3,874
4 authors:
Priyanka Madhushri Armen A Dzhagaryan

University of Alabama in Huntsville University of Alabama in Huntsville
13 PUBLICATIONS 34 CITATIONS 18 PUBLICATIONS 47 CITATIONS
SEE PROFILE SEE PROFILE
Emil Jovanov Aleksandar Milenkovic

University of Alabama in Huntsville University of Alabama in Huntsville
180 PUBLICATIONS 6,037 CITATIONS 114 PUBLICATIONS 3,654 CITATIONS
SEE PROFILE SEE PROFILE
Some of the authors of this publication are also working on these related projects:
• Application of biomedical engineering in preclinical and clinical practice – Grant ID III41007 View project
Optimization of Data Transfers Between Edge Devices and the Cloud Using Compression Utilities View project
All content following this page was uploaded by Priyanka Madhushri on 19 July 2016.
The user has requested enhancement of the downloaded file.

information
Article
An mHealth Tool Suite for Mobility Assessment
Priyanka Madhushri, Armen Dzhagaryan, Emil Jovanov and Aleksandar Milenkovic *
Electrical and Computer Engineering Department, The University of Alabama in Huntsville,
Huntsville, AL 35899, USA; pm0018@uah.edu (P.M.); aad0002@uah.edu (A.D.); emil.jovanov@uah.edu (E.J.)
* Correspondence: milenka@uah.edu; Tel.: +1-256-824-6830
Academic Editors: Jack Bobak Mortazavi, Hassan Ghasemzadeh, Sunghoon Ivan Lee and Nabil Alshurafa
Received: 1 April 2016; Accepted: 14 July 2016; Published: 18 July 2016
Abstract: The assessment of mobility and functional impairments in the elderly is important for
early detection and prevention of fall conditions. Falls create serious threats to health by causing
disabling fractures that reduce independence in the elderly. Moreover, they exert heavy economic
burdens on society due to high treatment costs. Modern smartphones enable the development
of innovative mobile health (mHealth) applications by integrating a growing number of inertial
and environmental sensors along with the ever-increasing data processing and communication
capabilities. Mobility assessment is one of the promising mHealth application domains. In this paper,
we introduce a suite of smartphone applications for assessing mobility in the elderly population.
The suite currently includes smartphone applications that automate and quantify the following
standardized medical tests for assessing mobility: Timed Up and Go (TUG), 30-Second Chair Stand
Test (30SCS), and 4-Stage Balance Test (4SBT). For each application, we describe its functionality and
a list of parameters extracted by processing signals from smartphone’s inertial sensors. The paper
shows the results from studies conducted on geriatric patients for TUG tests and from experiments
conducted in the laboratory on healthy subjects for 30SCS and 4SBT tests.
Keywords: mobility assessment; Timed Up and Go test; 30-Second Chair Stand Test; 4-Stage Balance
Test; inertial sensors; signal processing; health monitoring
1. Introduction
The quality of life of the elderly is highly correlated with their mobility. Reduced mobility creates
serious risk of fall. Falls in the elderly are defined as “unintentionally coming to the ground, or some
lower level because of sudden loss of balance, loss of consciousness, sudden onset of paralysis as in
stroke or an epileptic seizure” [1]. Falls are a major concern among the elderly because of their negative
impact on a person’s physical and physiological state. Falls are the third leading cause of worldwide
chronic disability [2], and a study found that approximately 81%–98% of hip fractures are caused
by falls in the elderly [3]. Falls not only create serious threats to health among the elderly, but the
treatment costs and resources required to address them also exert a heavy economic burden on the
society. The total costs resulting from falls were in the range of 0.85 to 1.5 percent of the total healthcare
costs in the United States, Australia, the United Kingdom and the European Union (EU) in 2009 [3].
These costs are expected to further increase, as there will be an estimated 88.5 million individuals
aged 65 and older in the United States alone in 2050, a 120% increase over the elderly population in
2010 [4]. The risks of falling and of fall related injury increase with the person’s age because of loss
of agility, vision loss, and medication side-effects [5]. In addition to physical damage and their high
costs, falls create fear of falling again by reducing self confidence among the elderly, particularly while
walking on uneven pathways or a wet floor. The consequences of falls also include increased risks
of pneumonia, pressure ulcers, and even death [6]. Therefore, the assessment of mobility and early
recognition of individuals prone to falls are the key strategies to reduce fall related injuries and prevent
their consequences.
Information 2016, 7, 47; doi:10.3390/info7030047 www.mdpi.com/journal/information

Information 2016, 7, 47 2 of 19
The Centers for Disease Control and Prevention (CDC) created the STEADI (Stopping Elderly
Accidents, Deaths & Injuries) tool kit for health care providers [7]. The CDC recommends the evaluation
of gait, strength, and balance using three tests: the Timed Up and Go (TUG) test [8], the 30-Second
Chair Stand Test (30SCS) [9], and the 4-Stage Balance Test (4SBT) [10]. These tests are generally
performed in a clinical setting, thus limiting the frequency of testing and increasing the total test
costs. Because of the demographic changes, there will be an increasing number of geriatric patients
in the near future and probably fewer professionals would be available to assess the risk of fall in
clinical settings. Therefore researchers have attempted to bring these assessment tests to patients’
homes [11]. In the new era of health monitoring, wireless sensing, mobile and cloud computing
technologies support the development of mobile health applications. These technologies have been
proven to effectively monitor the activities of daily living [12–14]. Automatic activity recognition
and quantification systems that utilize inertial sensors are proposed for long-term health and fitness
monitoring [15–17], assessment of mobility in the elderly and people with Parkinson’s disease [18,19],
automatic fall detection [20,21], and rehabilitation [22,23]. An instrumented Timed Up and Go (iTUG)
test has been recently introduced and proven to be sensitive to pathologies [24,25] and useful in fall
risk prediction [26]. Approaches for automatic activity recognition used by researchers vary in number,
type, and placement of utilized sensors, as well as in processing of recorded signals. While some
researchers used multiple sensors for automatic activity recognition [27–29], an increasing number of
projects use a single inertial sensor [30–33] usually placed on the subject’s chest. The rapid proliferation
of smartphones and continual growth in smartphone capabilities have opened up new opportunities
for health monitoring applications. Modern smartphones include a number of built-in inertial and
environmental sensors that can be utilized in health and fitness monitoring applications. Modern
smartphones typically include accelerometers, gyroscopes, magnetometers, barometers, and humidity
sensors. With growing data processing and communication capabilities, smartphones enable the
development of innovative mobile health (mHealth) applications.
In this paper, we introduce a suite of smartphone applications for mobility assessment. The suite
includes applications for automating and quantifying standard mobility tests recommended by the
Centers for Disease Control and Prevention (CDC): Timed Up and Go (TUG), 30-Second Chair Stand
Test (30SCS), and the 4-Stage Balance Test (4SBT). The names of the applications are sTUG (Smart TUG)
Doctor, 30SCS and 4SBT, respectively. The applications record and process the signals from the
smartphone’s accelerometer, gyroscope, and magnetometer sensors to extract the parameters that
quantify individual phases of the tests. The applications offer an affordable solution for quantifying
mobility of the elderly with an immediate feedback and automated logging. The test procedures
require minimum setup that includes a chair, a floor marker at the distance of three meters from the
chair, and an inexpensive instrumentation with a smartphone placed on the chest or belt running
the mobility assessment applications. The applications are quite easy to use and can be used in both
ambulatory and clinical settings. For example, in ambulatory settings elderly subjects can periodically
use applications to quantify their mobility and possibly detect early signs of mobility deterioration.
The tests can be self-guided or conducted with a help of caregiver. In clinical settings, healthcare
professionals can use applications to assess mobility of multiple patients. With automatic updates to
an mHealth server over the Internet and services provided by the mHealth web portal, healthcare
professionals can gain insights into mobility of tested subjects over time. For example, they can assess
the impact of therapeutic interventions, e.g., impact of drugs, by analyzing the parameters from
multiple tests performed in a single day. Next, healthcare professionals and researchers can monitor
and evaluate evolution of disease by analyzing the trends in the parameters collected over longer
periods of time.
The rest of the paper is organized as follows. Section 2 describes system architecture of the
mobility assessment framework. Section 3 describes TUG test procedure and application including
signal processing and parameter extraction. Sections 4 and 5 describe the 30SCS and 4SBT applications,
respectively, including parameters and signal processing. Section 6 describes the results from
applications, respectively, including parameters and signal processing. Section 6 describes the
results from preliminary
preliminary testson
tests performed performed on the
the geriatric geriatric
as well as well individuals.
as healthy as healthy individuals. Section 7
Section 7 concludes
concludes
the paper. the paper.
2.
2. Mobility
MobilityAssessment
AssessmentSystem
System
2.1. System
2.1. System Architecture
Architecture
Figure 11 illustrates
Figure illustrates system
system architecture
architecture of the application
application suite.
suite. The
The system
system requires
requires anan Android
Android
smartphone with built-in accelerometer, gyroscope and orientation sensors running
smartphone with built-in accelerometer, gyroscope and orientation sensors running Android 2.3 Android 2.3 or
or
above. During
above. During tests,
tests, the
the smartphone
smartphone isis mounted
mounted on on the
the subject’s
subject’s chest
chest or
or back
back using
using an
an elastic
elastic band
band
to keep
to keep itit stationary
stationary and and close
close to the body. The z-axis corresponds
The z-axis corresponds to to the
the sagittal
sagittal axis,
axis, the y-axis
the y-axis
corresponds to
corresponds to the
the longitudinal
longitudinal axis,
axis, and
and the x-axis corresponds
the x-axis corresponds to to the
the frontal
frontal axis
axis of
of the
the human
human
body. Smartphone applications
body. applications record
record and process signals from the sensors,
sensors, extract
extract parameters,
parameters, and and
display results on the smartphone’s
display results on the smartphone’s screen. The applications stop monitoring automatically
The applications stop monitoring automatically after the after the
completion of the test is detected. They create test descriptors that include date and
completion of the test is detected. They create test descriptors that include date and time when the time when the
tests are
tests are taken
taken and
and all
all parameters
parameters that
that quantify
quantify the
the tests.
tests. A
A test
test descriptor
descriptorisis stored
stored in
in aa .csv
.csv file
file on
on
the smartphone.
the smartphone.
Patient
Researchers /
Healthcare Professionals
X
Z Internet / Cloud
g
Y
mHealth Servers
Figure 1.
Figure System architecture
1. System architecture for
for mobility
mobility assessment.
assessment.
The
The smartphone
smartphone applications
applications can can be
be configured
configured for use in clinical settings. In In this
this mode
mode of of
operation, the applications require healthcare professionals to be registered
operation, the applications require healthcare professionals to be registered users in the mHealth server users in the mHealth
server databases
databases with privileges
with privileges to administer
to administer mobility mobility
tests. Atests. A test administrator
test administrator can enter cana enter a new
new patient
patient
or selector aselect a patient
patient already already registered
registered in the in system.
the system. Upon
Upon completionofofaatest,
completion test, the
the application
application
automatically
automatically uploads
uploads thethe test
testdescriptor
descriptortotothe themHealth
mHealthserver.server.The ThemHealth
mHealth server
serverconsists of
consists
multiple
of multiple virtual
virtualmachines
machines providing
providingstorage,
storage,database
databasemanagement,
management,access accesscontrol,
control, and and data
data
visualization
visualization services.
services. ItIt includes
includes three
three main
main components:
components: mHealth mHealthdatabase,
database,mHealth
mHealthWeb Web API,
API,
and
and mHealth
mHealth Web Web Portal [34]. The The mHealth
mHealth database
database is is built
built to to support
support aa variety
variety of physiological
physiological
records
records with
with annotations.
annotations. EachEach record
record hashas information
information about about thethe subject,
subject, equipment
equipment used used to to collect
collect
records,
records, and conditions under which the data are recorded, including application type, timestamp of
and conditions under which the data are recorded, including application type, timestamp of
the
the recording,
recording, and andlocation.
location.The TheWebWebAPI APIis is designed
designed to to
be be intermediary
intermediary between
between personal
personal devices
devices and
and the mHealth
the mHealth database.
database. Any interaction
Any interaction with thewith the mHealth
mHealth API requires APIuserrequires user authentication.
authentication. Authorized
Authorized
healthcare professionals and researchers can use the mHealth Web portal services for remote accessfor
healthcare professionals and researchers can use the mHealth Web portal services to
remote access to
physiological physiological
records and for records and for
rudimentary rudimentaryofvisualization
visualization data. Through of data.
theseThrough
services,these services,
longitudinal
longitudinal
studies aimed studies aimed at quantifying
at quantifying improvements improvements or deterioration
or deterioration in mobility incan
mobility can be performed.
be performed.
With
With thethe proposed
proposedmounting
mountingofofthe the smartphone
smartphone ononthethe
chestchest or back,
or back, the applications
the applications record record
the x,
the x, y,
y, and and z components
z components of the acceleration
of the acceleration of the subject’s
of the subject’s upper trunk upper andtrunk and use
use these these
signals signals to
to determine
determine
the upper bodythe upper body movement.
movement. The smartphone’s
The smartphone’s gyroscopeangular
gyroscope measures measures angular movements,
movements, specifically
specifically
the rotation the
aroundrotation around
the x-axis they-axis
(roll), x-axis (roll),and
(yaw), y-axis
z-axis(yaw),
(pitch).andIn z-axis (pitch).
our setup, In ourthe
we record setup, we
angular
record
velocitythe angular
from velocity from
the gyroscope and the
use gyroscope
the signalsand use the
to detect andsignals
locatetoimportant
detect andtransitions
locate important
during
transitions
the TUG and during
30SCSthetests.
TUGSmartphone’s
and 30SCS tests. Smartphone’s
magnetic sensor is magnetic
sensing the sensor
Earth’sis sensing
magnetic thefield
Earth’s
and
magnetic field and is normally used to aid navigation by determining the Earth’s magnetic poles and
is normally used to aid navigation by determining the Earth’s magnetic poles and the smartphone’s
orientation. In our setup,
the smartphone’s we useInthese
orientation. signals
our setup, wetouse
detect
thesethe subject’s
signals orientation
to detect in the
the subject’s TUG test.in
orientation
The
the TUGtests can be administered by the subject himself/herself or a healthcare practitioner in a
test.
The tests The
clinical setting. can tests
be administered
can be takenby many
the subject
timeshimself/herself or a healthcare
in a day. Analysis practitionercollected
of the parameters in a
overclinical
longersetting.
periodsTheoftests
timecan
maybe help
takeninmany timeslong-term
tracking in a day. Analysis
changes of in the parameters
mobility collected of
and diagnosis
over impairments.
balance longer periodsAllof the
timeapplications
may help inhave
tracking
beenlong-term
tested onchanges
Nexus 4insmartphone,
mobility andHTC
diagnosis
M7 and of M8
balance impairments. All the applications have been tested on Nexus 4 smartphone,
(HTC Corporation, Taoyuan, Taiwan), Motorola RAZR M, and RAZR HD (Motorola, Inc., Schaumburg, HTC M7 and
M8 (HTC Corporation, Taoyuan, Taiwan), Motorola RAZR M, and RAZR HD (Motorola, Inc.,
IL, USA).
Schaumburg, IL, USA).
2.2. Settings and Configurations
2.2. Settings and Configurations
Figure 2 shows the smartphone screens demonstrating the settings and configurations of the
Figure 2 shows the smartphone screens demonstrating the settings and configurations of the
applications when used in clinical settings. The applications need to be set up before conducting
applications when used in clinical settings. The applications need to be set up before conducting the
the tests
testsand
andthis
thisis is typically done by healthcare professionals in charge of administering the tests.
typically done by healthcare professionals in charge of administering the tests. First,
First,
four settings screens (a),(a),
four settings screens (b),(b),
(c) (c)
andand(d)(d)
areare common
common to all
to all thethe applications,
applications, whereas
whereas the the
lastlast
twotwo
settings screens
settings (e)(e)
screens and
and(f)(f)
areareused
usedforforthe
the30SCS
30SCS and
and sTUG Doctorapplications,
sTUG Doctor applications, respectively.
respectively.
(a) (b) (c)
(d) (e) (f)
Figure 2. Settings and configuration of the applications suite: (a) Login screen; (b) Configuration
Figure 2. Settings and configuration of the applications suite: (a) Login screen; (b) Configuration
screen; (c) Adding a new subject screen; (d) Subjects screen; (e) 30SCS configuration screen; (f) sTUG
screen; (c) Adding a new subject screen; (d) Subjects screen; (e) 30SCS configuration screen; (f) sTUG
Doctor configuration screen.
Doctor configuration screen.
Figure 2a shows a login interface. The applications use the Internet to communicate with the
mHealthFigure
server.2aThe applications
shows require application
a login interface. users,use
The applications typically healthcare
the Internet professionals,
to communicate withtothe
enter
their credentials
mHealth forThe
server. login to the mHealth
applications requireserver. A “setting”
application button on
users, typically the login
healthcare screen takestothe
professionals,
userenter their
to the credentials for
configuration loginshown
screen to the in
mHealth
Figure server.
2b. TheAcheckbox
“setting” “Remember
button on theMe” login screen the
enables takesuser
the user
to save andtocache
the configuration screen shown
the login credentials. Thein“Save
Figureto2b. The checkbox
External SD card”“Remember Me” enables
option enables saving the
raw
user to save and cache the login credentials. The “Save to External SD card”
data and tests descriptors to the external SD card memory. The “Offline Session” allows the user option enables saving
raw datatests
to perform and tests descriptors
without uploadingto thedata
external
on theSDserver.
card memory.
The “SaveThe Configurations”
“Offline Session” allows
optionthe userthe
saves
to perform
modified tests without
configuration. uploading
After loggingdataintoon theapplication,
the server. The it“Save
shows Configurations”
up a screen shownoptionin saves
Figurethe 2d
modified configuration. After logging into the application, it shows up a screen shown in Figure 2d
with the list of subjects (with subject ids and dates of birth) who already participated in the tests.
with the list of subjects (with subject ids and dates of birth) who already participated in the tests. The
The “add new subject” icon on the screen opens up a screen shown in Figure 2c and asks for the
“add new subject” icon on the screen opens up a screen shown in Figure 2c and asks for the new
new subject information such as unique id, gender and date of birth. The subject is added to the list
subject information such as unique id, gender and date of birth. The subject is added to the list of
of users when the “save” button on the screen is pressed. Screens shown in Figure 2e,f show the
users when the “save” button on the screen is pressed. Screens shown in Figure 2e,f show the
configurations
configurations available
availablefor
forthe
thesTUG
sTUGDoctor
Doctor and
and 30SCS
30SCS applications, respectively.“Upload
applications, respectively. “Upload TUG TUG
Parameters
Parameters to to
Server”
Server” enables
enablesthethedata
dataupload
upload to to the
the mHealth server.“Save
mHealth server. “SavetotoFiles”
Files” check
check boxes
boxes
enable the user to selectively save the raw data as well as parameters of the tests
enable the user to selectively save the raw data as well as parameters of the tests in the files. in the files. “Advance
Configuration” allows a user allows
“Advance Configuration” to specify the to
a user smartphone
specify thelocation, which
smartphone can be on
location, the subject’s
which can be onback the or
chest (default).
subject’s backForor research and development
chest (default). For research purposes, we allowed
and development options
purposes, we in the advanced
allowed options in settings
the
to set up sampling
advanced settingsfrequency and thresholds
to set up sampling frequencyused in thresholds
and the applications.
used in The sTUG DoctorThe
the applications. and 30SCS
sTUG
Doctor and
applications use30SCS applications
default thresholds useasdefault
shownthresholds
in Figureas 2f.shown in Figure 2f.
3. Smart Timed
3. Smart Up
Timed Upand
andGo
Go(sTUG
(sTUGDoctor)
Doctor) Application
Application
TheThe Timed
Timed UpUp andandGoGo (TUG)
(TUG) is aisfrequently
a frequently used
used clinical
clinical test
test forfor
thethe assessmentofofmobility,
assessment mobility,and
fall risk prediction in the elderly population and people with Parkinson’s disease [35]. This This
and fall risk prediction in the elderly population and people with Parkinson’s disease [35]. is a
is a simple
testsimple test to
and easy andadminister
easy to administer
anywhere, anywhere,
anytime. anytime.
In thisIntest,
thisatest, a subject
subject sitsa in
sits in a standard
standard armarm chair.
chair. The subject uses regular footwear and he/she can use a walking
The subject uses regular footwear and he/she can use a walking aid if needed. On the command “Go”, aid if needed. On the
the command “Go”,up
subject stands thefrom
subject
thestands up from
chair and walksthetochair and walks marker
a three-meter to a three-meter marker
on the floor at aon the floor
normal pace,
at a normal pace, turns around, walks back to the chair, and sits down again. Figure 3 demonstrates
turns around, walks back to the chair, and sits down again. Figure 3 demonstrates the typical phases
the typical phases of TUG test. The test outcome is the TUG time from the command “Go” to the
of TUG test. The test outcome is the TUG time from the command “Go” to the time when the subject
time when the subject sits on the chair again. An adult with mobility impairments may take more
sits on the chair again. An adult with mobility impairments may take more than 30 s to complete the
than 30 s to complete the test, whereas a healthy adult can perform this test in less than 10 s. This test
test, whereas a healthy adult can perform this test in less than 10 s. This test has been proven valuable
has been proven valuable in early assessment of balance and mobility, although it is limited as its
in early assessment
only outcome of time
is the balance and mobility,
to complete although
the test. An olderitadult
is limited as itsmore
who takes only than
outcome12 s tois complete
the time to
complete
the TUG test is at high risk for falling [36]. In addition, the health care professionals also observerisk
the test. An older adult who takes more than 12 s to complete the TUG test is at high andfor
falling [36].
report theIn addition,
subject’s the health
postural caregait,
stability, professionals
stride length,also
andobserve
sway toand fill report
in a formtheidentifying
subject’s postural
slow,
stability, gait,pace;
tentative stride length,
loss and sway
of balance; shorttostrides;
fill in alittle
formoridentifying
no arm swing; slow,shuffling;
tentativeor pace; loss ofuse
improper balance;
of
short strides;
walking aids.little or no arm swing; shuffling; or improper use of walking aids.
Figure 3. Timed Up and Go (TUG) test phases.

Figure 3. Timed Up and Go (TUG) test phases.
The sTUG Doctor application automates and quantifies the standard Timed Up and Go (TUG)
The
test sTUG
[37]. Doctor application
On application automates
start, it provides andto
options quantifies the standard
set up configuration andTimed
subjectUp and Go (TUG)
information as
testshown
[37]. On
in Figure 2. Then it opens the main display screen shown in Figure 4a. When the subject or as
application start, it provides options to set up configuration and subject information
shown in Figure 2. Then it opens the main display screen shown in Figure 4a. When the subject or
the test administrator presses the green “Start” button on the screen, the application starts recording
the test administrator presses the green “Start” button on the screen, the application starts recording
and processing data from in-built inertial sensors and calculates TUG test parameters in real time.
and processing data from in-built inertial sensors and calculates TUG test parameters in real time.
The application starts measuring time from the point it finds a sit-to-stand transition during the test.
The application starts measuring time from the point it finds a sit-to-stand transition during the test.
The application stops automatically after it detects a stand-to-sit transition and important
The application stops automatically after it detects a stand-to-sit transition and important parameters
parameters are calculated and displayed instantly on the screen as shown in Figure 4b. The
are calculated and displayed instantly on the screen as shown in Figure 4b. The application quantifies
application quantifies all the phases of the TUG test and calculates parameters listed in the Table 1.
all the phases of the TUG test and calculates parameters listed in the Table 1. These parameters
These parameters include the total TUG duration, sit-to-stand duration, stand-to-sit duration,
include the total TUG duration, sit-to-stand duration, stand-to-sit duration, maximum lean forward
maximum lean forward angle, maximum angular velocity, and the total number of steps taken
angle, maximum angular velocity, and the total number of steps taken during walking. The detailed
during walking. The detailed set of test parameters with subject’s unique id and timestamp are
set
saved on parameters
of test withand
the smartphone subject’s
sent tounique id andserver.
the mHealth timestamp are saved on
This application alsothe
hassmartphone and
an alternative
sent to the mHealth server. This application also has an alternative display tab which
display tab which keeps all the recordings for each subject with exact date and time of the test withkeeps all the
recordings
summary for each subject
parameters with exact
as shown date 4c.
in Figure andIt time of the
should test with
be noted thatsummary parameters
in standalone as shown
application mode,in
Figure 4c. It should be noted that in standalone application mode, subjects are expected
subjects are expected to be able to start the application and press the Start button. The screens to be able to
start
described in Figure 4a,b are designed to accommodate older adults by featuring big buttons andto
the application and press the Start button. The screens described in Figure 4a,b are designed
accommodate
large fonts witholder adults
posture by featuring big buttons and large fonts with posture icons.
icons.
(a) (b) (c)
Figure 4. sTUG Doctor smartphone application screens: (a) application screen before the test is
Figure 4. sTUG Doctor smartphone application screens: (a) application screen before the test is started;
started; (b) application screen with a selected set of parameters upon completion of a test; and
(b) application screen with a selected set of parameters upon completion of a test; and (c) application
(c) application screen showing a log with multiple measurements for a single subject.
screen showing a log with multiple measurements for a single subject.
Table 1. Parameters for TUG characterization.

Table 1. Parameters for TUG characterization.
Parameter Description Units
d.TUG
Parameter Total duration of the TUG test (from “Go” to the completion of the test)
Description sUnits
d.S2ST Total duration of the sit-to-stand transition; d.S2ST = d.LF + d.LT s
d.TUG Total duration of the TUG test (from “Go” to the completion of the test) s
d.LF Duration of the lean forward phase in the sit-to-stand transition s
d.S2ST Total duration of the sit-to-stand transition; d.S2ST = d.LF + d.LT s
d.LT Duration of the lift up phase in the sit-to-stand transition s
d.LF Duration of the lean forward phase in the sit-to-stand transition s
d.WALK Total time of walk s
d.LT Duration of the lift up phase in the sit-to-stand transition s
d.ST2S
d.WALK Duration of the stand-to-sit transition;
Total time of walk d.ST2S = d.PS + d.SD s s
d.PS
d.ST2S Duration of the
Duration prepare-to-sit
of the stand-to-sitphase in the d.ST2S
transition; stand-to-sit transition
= d.PS + d.SD s s
d.SD
d.PS Durationofofthe
Duration theprepare-to-sit
sit-down phase in the
phase in stand-to-sit
the stand-to-sittransition
transition s s
a.S2ST
d.SD Maximum
Durationchange of the trunk
of the sit-down angle
phase in in
thethe lean forward
stand-to-sit phase
transition degrees
s
v.LF
a.S2ST Maximumchange
Maximum angular ofvelocity
the trunkduring
anglethein lean forward
the lean phase
forward phase degrees/s
degrees
v.LF
v.LT Maximum
Maximum angular
angularvelocity during
velocity during the
thelean
lift forward
up phasephase degrees/s
degrees/s
v.LT
n.STEP Maximum angular
Total number velocity
of steps during
during the lift
walking up phase
phase degrees/s
steps
n.STEP
n.SBT TotalTotal
number of steps during walking
number of steps before turn phase steps
steps
n.SBT Total number of steps before turn steps
Figure 5 shows raw acceleration and gyroscope signals collected by the sTUG Doctor application
during Figure
a TUG 5test. When the “Start” button on the screen is pressed, the beginning timestamp is
shows raw acceleration and gyroscope signals collected by the sTUG Doctor
captured. The during
application program then test.
a TUG waitsWhenfor the
thebeginning
“Start” button of theonsit-to-stand
the screen transition
is pressed,by thesearching
beginningfor a
change in the angular velocity around the frontal axis (the x component
timestamp is captured. The program then waits for the beginning of the sit-to-stand transition of the gyroscope). This change
by
is caused
searching for a change in the angular velocity around the frontal axis (the x component thresholds
by leaning forward as the subject prepares to stand up. Additionally, acceleration of the
aregyroscope).
checked to ensure that the
This change is leaning
caused by forward
leaning signature
forwardisascaused by an actual
the subject prepares motion rather
to stand up.than
by swinging
Additionally,forward while still
acceleration in the are
thresholds sitting position.
checked The angular
to ensure that the velocity during the
leaning forward sit-to-stand
signature is
caused by an actual motion rather than by swinging forward while still
transition has a distinct profile as it starts from zero, increases to reach the maximum, and drops to in the sitting position. The
angular
a zero at thevelocity
end ofduring
the lean theforward
sit-to-stand
phase transition
(LF). To has a distinct
determine theprofile
beginningas it ofstarts from zero, the
the transition,
increases finds
application to reachthethe maximum,
maximum and drops
angular to a zero
velocity (v.LF)at the
thatend of the lean
is above forward
a certain phase (LF).
threshold, and Tothen
determine the beginning of the transition, the application finds the
searches backward to find the beginning of the slope. The maximum upper trunk angle (a.S2ST) maximum angular velocity (v.LF)
that is above a certain threshold, and then searches backward to find the beginning of the slope. The
is reached at the end of the lean forward phase. The time duration between the beginning of the
maximum upper trunk angle (a.S2ST) is reached at the end of the lean forward phase. The time
transition and the end of the lean forward phase represents the duration of the lean forward phase
duration between the beginning of the transition and the end of the lean forward phase represents
(d.LF). The second phase of the sit-to-stand transition is characterized by a negative angular velocity as
the duration of the lean forward phase (d.LF). The second phase of the sit-to-stand transition is
the characterized
subject movesbyinto an upright position. The angular velocity reaches the minimum (i.e., v.LT), and
a negative angular velocity as the subject moves into an upright position. The
then increases
angular back reaches
velocity to zero.theTheminimum
moment (i.e.,whenv.LT),it becomes
and then positive is considered
increases back to zero. to be themoment
The end of the
lift when
up phaseit becomes positive is considered to be the end of the lift up phase and the end of the the
and the end of the sit-to-stand transition. By time stamping this moment, we calculate
duration of thetransition.
sit-to-stand lift up phaseBy (d.LT), and thethis
time stamping totalmoment,
durationwe ofcalculate
sit-to-standthe transition,
duration ofd.S2ST
the lift =up
d.LF
phase+ d.LT.
A stand-to-sit
(d.LT), and thetransition can also
total duration be divided
of sit-to-stand into twod.S2ST
transition, separate= d.LFphases,
+ d.LT.aAprepare-to-sit (PS) and a
stand-to-sit transition
sit-down
can also(SD) phase. into
be divided The two
angular velocity
separate phases, increases to the maximum
a prepare-to-sit (PS) and aand then drops
sit-down back The
(SD) phase. to zero
angular velocity increases to the maximum and then drops back
as the subject leans forward in preparation to sit down. The moment when the angular velocityto zero as the subject leans forward
dropsin preparation
to zero marks to sitthe
down.
endThe of themoment when the angular
prepare-to-sit phase and velocity drops to zero
the beginning of marks the end phase.
the sit-down of
the prepare-to-sit
By timestamping thesephase and thepoints,
characteristic beginning of the sit-down
the application determinesphase.the By timestamping
duration these
of the preparation
characteristic points, the application determines the duration of the preparation
to sit phase (d.PS). In the sit-down phase, the angular velocity is negative as the subject’s upper to sit phase (d.PS).
In the sit-down phase, the angular velocity is negative as the subject’s upper trunk moves back into
trunk moves back into the upright position. The moment it becomes positive marks the end of the
the upright position. The moment it becomes positive marks the end of the sit-down phase and the
sit-down phase and the entire stand-to-sit transition. By capturing the time stamp of this moment the
entire stand-to-sit transition. By capturing the time stamp of this moment the application calculates the
application calculates the duration of the sit-down phase (d.SD), and the entire stand-to-sit transition
duration of the sit-down phase (d.SD), and the entire stand-to-sit transition (d.ST2S = d.PS + d.SD). The
(d.ST2S = d.PS + d.SD). The total walk time (d.WALK) is calculated using captured timestamps and
total walk time (d.WALK) is calculated using captured timestamps and the number of steps
the(n.STEP)
number using
of stepsthe(n.STEP) using
acceleration and the acceleration
gyroscope and gyroscope
thresholds. thresholds.
The application uses The application uses
the smartphone’s
themagnetic
smartphone’s magnetic sensor data to detect turning and the
sensor data to detect turning and the number of steps before turn (n.SBT). number of steps before turn (n.SBT).
(a)
(b)
Figure 5. Inertial signals recorded during a TUG test: (a) raw signals from accelerometer, (b) raw
Figure 5. Inertial signals recorded during a TUG test: (a) raw signals from accelerometer; (b) raw signals
signals from gyroscope.
from gyroscope.
Information
4. The 2016, 7, 47
30-Second Chair Stand (30SCS) Application 8 of 20
The
4. The30-Second
30-SecondChair
Chair Stand (30SCS)
Test measures the number of stand ups a person can perform during
Application
a 30 s interval
The 30-Second Chair Stand Test measures the number test
which is based on the five-times-sit-to-stand [38]. ups
of stand The aprimary
person cangoalperform
of this test is
to measure
during athe 30 slower extremity
interval strength
which is based andfive-times-sit-to-stand
on the endurance, but it can testalso
[38]. indicate
The primaryspeed,
goalbalance,
of this and
mobility [39–41].
test is The standard
to measure the lowertest is limited
extremity as its and
strength onlyendurance,
quantitativebutoutcome is the
it can also number
indicate speed,of stands
balance,within
completed and mobility [39–41].
30 s. The testThe standard test
is conducted is limited
using as its back
a straight only quantitative
chair without outcome is the and a
arm rests,
number The
stopwatch. of stands completed
patient within
is sitting in the30 middle
s. The testofisthe
conducted using
chair with a straight
feet flat on back chair without
the floor, hands placed
on opposite shoulders and crossed at the wrists. The typical phases of 30SCS test on
arm rests, and a stopwatch. The patient is sitting in the middle of the chair with feet flat arethe floor, in the
shown
hands placed on opposite shoulders and crossed at the wrists. The typical phases of 30SCS test are
Figure 6. On the command “Start” the patient rises to a full standing position and then sits back down
shown in the Figure 6. On the command “Start” the patient rises to a full standing position and then
holding arms against his/her chest. The patient keeps repeating these steps for 30 s. Table 2 gives
sits back down holding arms against his/her chest. The patient keeps repeating these steps for 30 s.
30SCS test 2scores
Table gives as a function
30SCS of gender
test scores and ageofused
as a function gender to determine
and age used whether the patient
to determine is atthe
whether high risk
for falls [42].
patient is A below
at high average
risk for fallsscores
[42]. A indicates a high
below average riskindicates
scores of falls. a high risk of falls.
Figure 6. The 30-Second Chair Stand (30SCS) test phases.

Figure 6. The 30-Second Chair Stand (30SCS) test phases.
Table 2. The 30-Second Chair Stand Test scores with ages and genders.
Table 2. The 30-Second Chair Stand Test scores with ages and genders.
Men Women
Age
Below Avg. Average Men Above Avg. Below Avg. Average
Women Above Avg.
Age
60–64 <14 14–19 >19 <12 12–17 >17
Below Avg. Average Above Avg. Below Avg. Average Above Avg.
65–69 <12 12–18 >18 <11 11–16 >16
60–64
70–74 <14
<12 14–19
12–17 >19
>17 <10<12 10–1512–17 >15 >17
65–69
75–79 <12
<11 12–18
11–17 >18
>17 <10<11 10–1511–16 >15 >16
70–74 <12 12–17 >17 <10 10–15 >15
80–84 <10 10–15 >15 <9 9–14 >14
75–79 <11 11–17 >17 <10 10–15 >15
85–89 <8 8–14 >14 <8 8–13 >13
80–84 <10 10–15 >15 <9 9–14 >14
90–94
85–89 <7
<8 7–12
8–14 >12
>14 <4 <8 4–11 8–13 >11 >13
90–94 <7 7–12 >12 <4 4–11 >11
The 30SCS smartphone application automates and quantifies the standard 30-Second Chair
Stand Test. Application begins the test when a subject presses the green “Start” button on the screen
The 30SCS
shown smartphone
in Figure 7a. The application
subject then automates
makes as many and quantifies
stand-ups asthehe/she
standard
can 30-Second
during 30 s.Chair
The Stand
Test. application
Application begins
makes anthe test when
acoustic a subject
cue (beep presses
sound) afterthe
30 green “Start” the
s to indicate button
test on the screenOur
completion. shown in
application counts the total number of stand-ups as well as total number of
Figure 7a. The subject then makes as many stand-ups as he/she can during 30 s. The application makes complete cycles
(stand-up
an acoustic cueto(beep
sit-down).
sound) The complete
after 30 s to set of derived
indicate parameters
the test for each
completion. cycle of 30SCS
Our application test the
counts is total
presented in Table 3. The application displays important parameters immediately on the screen after
number of stand-ups as well as total number of complete cycles (stand-up to sit-down). The complete
the completion of the test as shown in Figure 7b. The application also has an alternative tab to
set of derived parameters for each cycle of 30SCS test is presented in Table 3. The application displays
display the test summary with important test parameters for a single subject with data and time as
important parameters immediately on the screen after the completion of the test as shown in Figure 7b.
The application also has an alternative tab to display the test summary with important test parameters
for a single subject with data and time as shown in Figure 7c. The values of parameters listed along
with the raw signals of the test are saved in the .csv file on the smartphone as well as uploaded on the
mHealth server.
shown in2016,
Information Figure
7, 477c.The values of parameters listed along with the raw signals of the test are saved
9 of 19
in the .csv file on the smartphone as well as uploaded on the mHealth server.
(a) (b) (c)
Figure 7. 30SCS smartphone application screens: (a) application screen before the test is started; (b)
Figure 7. 30SCS smartphone application screens: (a) application screen before the test is started;
application screen with a selected set of parameters upon completion of a test; and (c) application
(b) application screen with a selected set of parameters upon completion of a test; and (c) application
Table 3.
Table 3. Parameters
Parameters for
for 30SCS
30SCS characterization.
characterization.

Parameter
n.SUP Total number Description
of stand-ups in 30 s Units-
n.SUP
n.CC Total number
Total number of ofstand-ups
completeincycles30 s - -
d.CCn.CC Total Total numberofofall
duration complete
complete cycles
cycles - s
d.CC Total duration of all complete cycles s
d.CYCLi
d.CYCLi
Duration of each cycle (d.CYCLi = d.S2STi + d.ST2Si + d.STi + d.SIT)
Duration of each cycle (d.CYCLi = d.S2STi + d.ST2Si + d.STi + d.SIT) s
s
d.S2STi
d.S2STi Total duration of the i ththsit-to-stand transition: d.S2STi = d.LFi + d.LTi
Total duration of the i sit-to-stand transition: d.S2STi = d.LFi + d.LTi s s
d.LFi
d.LFi Duration
Duration of of
thethelean
leanforward
forward phase
phase in inthe
thesit-to-stand
sit-to-stand transition
transition s s
d.LTi
d.LTi Duration
Duration ofofthetheliftliftup
upphase
phase inin the
thesit-to-stand
sit-to-stand transition
transition s s
d.ST2Si
d.ST2Si Duration of the stand-to-sit transition; d.ST2S
Duration of the stand-to-sit transition; d.ST2Si = d.PSi + d.SDi i = d.PS i + d.SD i s s
d.PSi Duration of the prepare-to-sit phase in the stand-to-sit transition s
d.PSi
d.SDi Duration of the
Duration of prepare-to-sit
the sit-down phase phase in the
in the stand-to-sit
stand-to-sit transition
transition s s
d.SDi
d.STi Duration of theTotal sit-down phase
duration in the stand-to-sit
of standing phase transition s s
d.STi
d.SIT Total
Total duration
durationofofstanding
sitting phasephase s s
a.S2STi
d.SIT Maximum change of the trunk angle
Total duration of sitting phase in the lean forward phase degrees s
v.LFi Maximum angular velocity during the lean forward phase degrees/s
a.S2STi Maximum change of the trunk angle in the lean forward phase degrees
v.LTi Maximum angular velocity during the lift up phase degrees/s
v.LFi Maximum angular velocity during the lean forward phase degrees/s
v.LTi Maximum angular velocity during the lift up phase degrees/s
Raw acceleration and gyroscope data for three chair stands during a 30SCS test is presented in
Figure 8. The
Raw plots are annotated
acceleration and gyroscope with markers
data forindicating
three chairrelevant
stands events
duringfor each cycle
a 30SCS test of
is apresented
chair stand. in
The test starts when the command “Start” button is pressed. Similar to
Figure 8. The plots are annotated with markers indicating relevant events for each cycle of a chairthe sTUG application, the 30SCS
application
stand. The testdetects a stand-up
starts when thephase command with its sub-phases
“Start” button(leaning forward
is pressed. Similarandto lifting up). Aapplication,
the sTUG subject can
spend some time in the standing position, described by a parameter
the 30SCS application detects a stand-up phase with its sub-phases (leaning forward and lifting d.STi. The standing is followed
up).
by a stand-to-sit transition. In the sitting position, a subject may
A subject can spend some time in the standing position, described by a parameter d.STi. The also stay for some time before a new
cycle is started.
standing The sitting
is followed is characterized
by a stand-to-sit for each
transition. cycle
In the by a parameter
sitting position, a d.SITi.
subjectPlease
may alsonotestaythatforin
healthy subjects the parameters d.STi and d.SITi may be equal to zero,
some time before a new cycle is started. The sitting is characterized for each cycle by a parameteri.e., the subjects may move from
one transition
d.SITi. to thethat
Please note other withoutsubjects
in healthy any delays. However, elderly
the parameters d.STi andpeople andmay
d.SITi people with balance
be equal to zero, and i.e.,
mobility impairments may require some time to prepare for the next
the subjects may move from one transition to the other without any delays. However, elderly people cycle. To detect characteristic
events,
and peoplethe starting and ending
with balance points impairments
and mobility of transitions,may algorithms used in
require some timethetosTUG application
prepare for the next are
slightly modified to work for the 30SCS test. The signals from the gyroscope sensors are processed to
cycle. To detect
Information characteristic events, the starting and ending points of transitions, algorithms10used
2016, 7, 47 of 19
in the sTUG application are slightly modified to work for the 30SCS test. The signals from the
gyroscope sensors are processed to timestamp all the phases of each cycle for 30SCS. The application
timestamp
counts all the phases
the number of each cycle
of stand-ups, and for
the 30SCS.
numberTheof application counts the numbercycles
complete stand-up/sit-down of stand-ups,
during and
the
the
test. number of complete stand-up/sit-down cycles during the test.
(a)
(b)
Figure 8. Inertial
Figure 8. Inertialsignals
signals recorded
recorded during
during three three
cycles cycles of test:
of a 30SCS a 30SCS test:
(a) raw (a) from
signals raw accelerometer;
signals from
accelerometer;
(b) raw signals(b) raw
from signals from gyroscope.
gyroscope.
5. The 4-Stage Balance Test (4SBT) Application

5. The 4-Stage Balance Test (4SBT) Application
The 4-stage balance test is used to assess static balance of the elderly during four progressively
The 4-stage balance test is used to assess static balance of the elderly during four progressively
more challenging balancing positions. In this test, the subject should not use any assistive devices
more challenging balancing positions. In this test, the subject should not use any assistive devices and
and should keep their eyes open during the test. The four respective positions of the 4SBT are shown
should keep their eyes open during the test. The four respective positions of the 4SBT are shown in
in Figure 9. The positions are described as follows: (a) Feet together stand assumes feet are placed side
Figure 9. The positions are described as follows: (a) Feet together stand assumes feet are placed side by
by side; (b) Semi-tandem stand assumes that the instep of one foot is touching the big toe of the other
side; (b) Semi-tandem stand assumes that the instep of one foot is touching the big toe of the other foot;
foot; (c) Tandem stand assumes that one foot is placed in front of the other-heel of one foot touches the
(c) Tandem stand assumes that one foot is placed in front of the other-heel of one foot touches the toe
toe of other foot; and (d) One leg stand assumes standing on one leg. A nurse supervising the test
of other foot; and (d) One leg stand assumes standing on one leg. A nurse supervising the test clearly
clearly describes and demonstrates each position and always stands next to the subject to hold
describes and demonstrates each position and always stands next to the subject to hold his/her arm
his/her arm and help them assume the correct foot position. When the subject is steady, the nurse
and help them assume the correct foot position. When the subject is steady, the nurse starts measuring
starts measuring time using a stopwatch, but remains ready to catch the subject in case of a balance
time using a stopwatch, but remains ready to catch the subject in case of a balance loss. If the subject
loss. If the subject can hold a position for 10 s without moving his/her feet or needing support, the
can hold a position for 10 s without moving his/her feet or needing support, the test advances to the
test advances to the next position. The standard test outcomes are times in seconds the subject is able
next position. The standard test outcomes are times in seconds the subject is able to hold each of four
to hold each of four positions. It has been shown that an older adult who cannot hold the tandem
positions. It has been shown that an older adult who cannot hold the tandem stance for at least 10 s is
stance for at least 10 s is at increased risk of falling [10].
at increased risk of falling [10].
The 4SBT smartphone application consists of four tests that can be selected using the test selector
screen shown in Figure 10a. The application starts the test after the green “Start” button on the display
screen is pressed as shown in Figure 10b. The standard outcome of this test is the number of seconds a
subject can hold the position. We have quantified the subject’s ability to hold the balance by calculating
the displacement of the chest and timestamping excessive movements. Table 4 describes the derived
4SBT test parameters. The 4SBT application provides the average chest movements during each of
the four phases/positions of the balance test. The application uses accelerometer signals to quantify
relative displacement of the chest. The chest displacement values indicate the balance control of
the subject; the lower displacement values demonstrate better stability during the tests. When the
subject stumbles during any of the four tests and comes back in the stable position (e.g., put the foot
down during one leg stand), the 4SBT application reports the time of stumbling and loss of balance.
Figure 10c 2016,
Information presents
7, 47 the display of parameters as well as the stumble status. 11 of 20
Figure 9. The 4-Stage Balance Test (4SBT) feet positions.
The 4SBT smartphone application consists of four tests that can be selected using the test
selector screen shown in Figure 10a. The application starts the test after the green “Start” button on
the display screen is pressed as shown in Figure 10b. The standard outcome of this test is the number
of seconds a subject can hold the position. We have quantified the subject’s ability to hold the
balance by calculating the displacement of the chest and timestamping excessive movements. Table 4
describes the derived 4SBT test parameters. The 4SBT application provides the average chest
movements during each of the four phases/positions of the balance test. The application uses
accelerometer signals to quantify relative displacement of the chest. The chest displacement values
indicate the balance control of the subject; the lower displacement values demonstrate better
stability during the tests. When the subject stumbles during any of the four tests and comes back in
the stable position (e.g., put the foot down during one leg stand), the 4SBT application reports the
time of stumbling and loss of balance.
Figure 9. The FigureBalance
The 4-Stage
4-Stage 10c presents the display of parameters as well as the
Figure 9. Balance Test
Test (4SBT)
(4SBT) feet
feetpositions.
positions.
stumble status.
The 4SBT smartphone application consists of four tests that can be selected using the test
selector screen shown in Figure 10a. The application starts the test after the green “Start” button on
the display screen is pressed as shown in Figure 10b. The standard outcome of this test is the number
of seconds a subject can hold the position. We have quantified the subject’s ability to hold the
balance by calculating the displacement of the chest and timestamping excessive movements. Table 4
describes the derived 4SBT test parameters. The 4SBT application provides the average chest
movements during each of the four phases/positions of the balance test. The application uses
accelerometer signals to quantify relative displacement of the chest. The chest displacement values
indicate the balance control of the subject; the lower displacement values demonstrate better
stability during the tests. When the subject stumbles during any of the four tests and comes back in
the stable position (e.g., put the foot down during one leg stand), the 4SBT application reports the
time of stumbling and loss of balance. Figure 10c presents the display of parameters as well as the
stumble status.
(a) (b) (c)
Figure 10. 4SBT smartphone application screens: (a) application screen for test position selection; (b)
Figure 10. 4SBT smartphone application screens: (a) application screen for test position selection;
application screen before the test is started; and (c) application screen with a selected set of
(b) application screen before the test is started; and (c) application screen with a selected set of
parameters upon completion of a test.
Table 4. Parameters for 4-Stage Balance Test characterization.

s.FTSi Relative displacement of the chest every second in the feet together stand cm
s.STSi Relative displacement of the chest every second in the semi-tandem stand cm
s.TSi Relative displacement of the chest every second in the tandem stand cm
s.OLi Relative displacement of the chest every second in the one leg stand cm
(a) (b) (c)

Figure 11 shows the chest acceleration magnitude during a semi-tandem stand when a healthy
subjectFigure 10. 4SBT
maintains smartphone
stable positionapplication screens:
(top graph), (a) application
maintains screen forwith
stable position test position selection;
an effort (middle (b)graph),
application screen before the test is started; and (c) application screen with a selected set of
and when the balance is lost approximately after two seconds from the beginning of the test (bottom
s.TSi Relative displacement of the chest every second in the tandem stand cm
s.OLi Relative displacement of the chest every second in the one leg stand cm
Figure 11 shows the chest acceleration magnitude during a semi-tandem stand when a healthy
subject maintains stable position (top graph), maintains stable position with an effort (middle
Information 2016, 7, 47
graph), and when the balance is lost approximately after two seconds from the beginning of the 12 of 19
test
(bottom graph). The graphs show that acceleration magnitude can indeed be used as a metric for
quantifying stability; significant changes in the acceleration magnitude in the bottom graph are due
graph). The graphs show that acceleration magnitude can indeed be used as a metric for quantifying
to a loss of balance. The 4SBT application calculates the displacement for each second of the test
stability; significant changes in the acceleration magnitude in the bottom graph are due to a loss
using Equations (1)–(4). Acc is the raw 3D acceleration magnitude (Equation (1)), Acc’ represent Acc
of balance. The 4SBT application calculates the displacement for each second of the test using
without the DC offset, eliminated by subtracting the value of Acc at rest, as shown in Equation (2).
Equations (1)–(4). Acc is the raw 3D acceleration magnitude (Equation (1)), Acc’ represent Acc without
The velocity vel is calculated at every sample using Equation (3), where ∆t is the sampling interval
the DC offset, eliminated by subtracting the value of Acc at rest, as shown in Equation (2). The velocity
(10 ms in our case). Next, the displacement Disp is calculated as shown in Equation (4). After
vel is calculated at every sample using Equation (3), where ∆t is the sampling interval (10 ms in
analyzing the displacement data from a several tests performed in the laboratory with and without
our case). Next,
stumbling, we havethe displacement Disp is calculated
found a displacement threshold of as30shown
cm thatin reliably
Equation (4). After
detects analyzing
the stumbling forthe
displacement
the existing data
data from a several tests
sets However, moreperformed in theinlaboratory
tests collected with and
clinical settings arewithout
necessary stumbling,
to find awe
have found a displacement threshold of 30 cm that reliably detects the stumbling
threshold that would work reasonably well in majority of test cases. It should be noted that the for the existing data
sets However, more tests collected in clinical settings are necessary to find
current algorithm for detecting a stumbling could fail if a subject drops a foot gently while a threshold that would work
reasonably
maintaining well in majority
a perfect of testAn
stability. cases. It should
alternative to be notedon
relying that the current
algorithms to algorithm for detecting
detect stumbling is to a
stumbling could fail if a subject drops a foot gently while maintaining a perfect
allow test administrator to automatically annotate a stumbling event when it occurs (e.g., by stability. An alternative
topressing
relying on algorithms
a button to detect
or issuing stumbling
a voice command). is to allow test administrator to automatically annotate a
stumbling event when it occurs (e.g., by pressing a button or issuing a voice command).
Acc  Accx 2  Acc y 2  Accz 2 (1)
b
2 2 2
Acc “ Acc x ` Accy ` Accz (1)
Acc '  Acc  Acc(rest) (2)
Acc1 “ Acc ´ Accprestq (2)
vel(i)  vel(i  1)  Acc '(i ) * t (3)
velpiq “ velpi ´ 1q ` Acc1piq ˚ ∆t (3)
1 1 (4) (4)
Disp“
Disppiq  Disp(´
(i )Disppi i 1q  vel
1)` (i ´1)1q
velpi * ˚t ∆t
 `Acc '(i) * t 2˚ ∆t2
Acc1piq
2 2
(a)
(b)
(c)
Figure 11. Comparison of accelerometer vector magnitudes during semi-tandem stands for:
(a) stationary standing; (b) standing with balance trail; and (c) standing with stumble.
6. Application Suite Verification and Validation

In this section we briefly discuss our experiments aimed at verification of the applications in the
mobility assessment suite. During the development of the algorithms for posture transitions from
mobility assessment suite. During the development of the algorithms for posture transitions from the
the smartphone’s inertial sensors in the TUG and 30SCS tests we used the following approach. A
number of volunteers participated in data collection, ranging from healthy subjects that performed
the tests as quickly as possible to people with mobility impairments affiliated with a local
Parkinson’s support group. The inertial sensor data collected by the smartphone are used in the
initial development of offline signal processing algorithms in Matlab (R1024a, The MathWorks, Inc.,
Natick, MA, USA). To guide and inform algorithm development, each session is videotaped using a
(c)

mobility assessment suite. During the development of the algorithms for posture transitions from
smartphone’s inertialinertial
the smartphone’s sensors in the in
sensors TUGthe and
TUG30SCS tests we
and 30SCS used
tests wethe
usedfollowing approach.
the following A number
approach. A
of number
volunteers participated in data collection, ranging from healthy subjects that
of volunteers participated in data collection, ranging from healthy subjects that performedperformed the tests as
quickly as possible
the tests to people
as quickly with mobility
as possible to people impairments
with mobility affiliated with a local
impairments Parkinson’s
affiliated with a support
local
group. The inertial
Parkinson’s sensor
support data
group. Thecollected
inertial by the smartphone
sensor data collectedare byused in the initialare
the smartphone development
used in the of
offline
initialsignal processing
development algorithms
of offline in Matlab algorithms
signal processing (R1024a, The MathWorks,
in Matlab (R1024a,Inc., The Natick, MA,Inc.,
MathWorks, USA).
To Natick,
guide andMA, inform
USA). To guide anddevelopment,
algorithm inform algorithm eachdevelopment, each sessionusing
session is videotaped is videotaped
a regular using a
camera
andregular
videoscamera and videos
are manually are manually
annotated annotated by
by timestamping timestamping
posture transitions. posture
For atransitions. For a of
selected group
selected group
participants we alsoof participants
performed the wetest
alsoinperformed
a laboratory theequipped
test in a laboratory
with the Vicon equipped
motionwith the Vicon
capture system.
motion capture system. This laboratory allows for accurate tracking of reflective
This laboratory allows for accurate tracking of reflective markers using 33 Vicon T40 series IR markers using 33
cameras
Vicon
(Vicon T40 seriesInc.,
Industries, IR cameras
Ipswich,(Vicon Industries,
UK). Test Inc.,the
users carry Ipswich, UK). Test
smartphone andusers carry the
strategically smartphone
placed reflective
and strategically placed reflective markers as shown in Figure 12. The
markers as shown in Figure 12. The motion capture system records the absolute position of each motion capture system
records
marker theaabsolute
with millimeter position of each
precision everymarker
20 ms. with a millimeter
Using these data, precision
we extract everythe20acceleration
ms. Using these
vectors
data, we extract the acceleration vectors from two reflective markers attached to smartphone and
from two reflective markers attached to smartphone and compare them to the acceleration vector
compare them to the acceleration vector recorded synchronously by the smartphone during typical
recorded synchronously by the smartphone during typical posture transitions. Whereas the algorithms
posture transitions. Whereas the algorithms for detection and characterization of posture transitions
for detection and characterization of posture transitions used in the smartphone applications rely
used in the smartphone applications rely exclusively on the sensor data from the smartphone, the
exclusively on the sensor data from the smartphone, the Vicon system provides absolute location that is
Vicon system provides absolute location that is used for verification of algorithms. We find that
used for verification
accelerometer dataofcaptured
algorithms. We find
by the that accelerometer
smartphone match the data captured by
accelerometer datathederived
smartphone
from match
the
thereflective
accelerometer data derived from the reflective markers attached to the
markers attached to the smartphone. In addition, we verify that angles and angular smartphone. In addition,
wevelocity
verify that angles
signals and angular
calculated by thevelocity signals
smartphone calculatedmatch
application by thethose
smartphone
derived application
from the Viconmatch
those
system—the difference between the two is less than 1%. Once the algorithms were tested andthe
derived from the Vicon system—the difference between the two is less than 1%. Once
algorithms
perfectedwere testedthey
in Matlab, andwere
perfected
portedintoMatlab,
Java and they
the were
Androidported
OS. to Java and the Android OS.
Figure 12. A subject instrumented by reflective markers and the smartphone during posture
Figure 12. A subject instrumented by reflective markers and the smartphone during posture detection
detection algorithms verification.
algorithms verification.
The second phase of the testing is carried out in both laboratory conditions and in clinical settings
under supervision of healthcare professionals. In the laboratory conditions, a variety of subjects
perform the mobility assessment tests. We capture videos of these sessions. These videos are manually
annotated to identify timestamps of posture transitions. Time parameters describing duration of
posture transitions are extracted from videos and compared to the time parameters reported by the
sTUG and 30SCS applications. Below we show the results of verification for a selected set of parameters.
Table 5 shows the results of a laboratory testing of the sTUG Doctor application, specifically
the verification of the dTUG parameter, which is the main quantitative outcome of the TUG test.
We consider three subjects, each performing three TUG tests (Test#1, Test#2, and Test#3). For each
test we show the time reported by the sTUG application (sTUG) and the time extracted from the
video, as well as the absolute error. The rows marked as Error indicate the absolute difference in
measurements, if we assume the times extracted from the video are taken as reference times. The results
show that the maximum error falls below 4%. In a clinical setting, the sTUG application is tested as
follows. A nurse in charge of mobility assessment carries out the standard TUG test with a stopwatch
to measure the time a subject needs to complete a TUG test. In addition, the subject has a smartphone
on his or her chest or back and the start button is pressed when the command GO is issued by the
nurse. The time recorded by the nurse using the stopwatch is compared to the time reported by the
sTUG Doctor application. In an experiment with 20 older adults we find that the test times recorded
by the stopwatch with the resolution of 0.1 s match the times reported by the sTUG in all conducted
tests within 400 ms.
Table 5. Testing the sTUG Doctor application: duration of the TUG test (d.TUG).
- - Test#1 Test#2 Test#3

sTUG (s) 10.6 9.84 9.54
S#1 (Female, 28) Video (s) 10.2 9.69 9.59
Error (%) 3.92 1.55 0.52
sTUG (s) 10.5 10.5 9.8
S#2 (Male, 47) Video (s) 10.38 10.54 9.76
Error (%) 1.16 0.38 0.41
sTUG (s) 8.79 8.61 8.48
S#3 (Male, 55) Video (s) 8.57 8.47 8.64
Error (%) 2.57 1.65 1.85
Table 6 shows the results of a laboratory testing of the 30SCS application. We consider three
subjects S#1, S#2, and S#3, each performing the test two times (Test#1 and Test#2). For each test,
we show the times to perform the sit-to-stand transitions. The rows marked with 30SCS include
the sit-to-stand times in seconds reported for individual stand-ups during the test reported by the
application, and rows marked with Video include the sit-to-stand times extracted from manually
annotated videos. If we assume that times extracted from the videos are reference times, the absolute
errors by the 30SCS application range from 0% to 14.4%, but are typically less than 5%. We find these
errors quite satisfactory. For the medical outcome of the 30SCS test, the number of stand-ups, the
30SCS application is 100% accurate. Similar observations can be found for the stand-to-sit times—the
absolute errors for the times reported by the 30SCS range between 0% and 3.7%.
Table 6. Testing the 30SCS application: sit-to-stand transition times (d.S2STi).
S#1 (Female, 28) - #1 #2 #3 #4 #5 #6 #7 #8 #9 #10 #11 #12

30SCS (s) 1.37 0.76 0.77 0.78 0.77 0.8 0.84 0.79 0.83 0.83 0.84 0.84
Test #1 Video (s) 1.5 0.75 0.75 0.75 0.75 0.8 0.83 0.78 0.85 0.8 0.83 0.83
Error (%) 14.38 1.33 2.67 4.00 2.67 0.00 1.20 1.28 2.35 3.75 1.20 1.20
30SCS (s) 1.18 0.92 0.95 1 0.92 0.98 0.91 0.99 1 - - -
Test #2 Video (s) 1.25 0.9 0.9 1.02 0.9 0.95 0.9 0.93 1.02 - - -
Error (%) 5.60 2.22 5.56 1.96 2.22 3.16 1.11 6.45 1.96 - - -
S#2 (Male, 47) - #1 #2 #3 #4 #5 #6 #7 #8 #9 #10 #11 -
30SCS (s) 1.46 1.06 1.09 1.07 1.04 1.02 1.04 1.05 1.04 1.05 1.06 -
Test #1 Video (s) 1.48 1.08 1.05 1.11 1.06 1.04 1.05 1.09 1.03 1.02 1.01 -
Error (%) 1.35 1.85 3.81 3.60 1.89 1.92 0.95 3.67 0.97 2.94 4.95 -
30SCS (s) 0.98 1.2 1.54 1.54 1.3 1.22 1.42 - - - - -
Test #2 Video (s) 1.01 1.2 1.54 1.57 1.26 1.21 1.44 - - - - -
Error (%) 2.97 0.00 0.00 1.91 3.17 0.83 1.39 - - - - -
S#3 (Male, 28) - #1 #2 #3 #4 #5 #6 #7 #8 #9 #10 - -
30SCS (s) 1.61 1.03 1.19 1.2 1.19 1.22 1.23 1.27 1.2 1.24 - -
Test #1 Video (s) 1.6 0.98 1.22 1.24 1.18 1.2 1.28 1.26 1.18 1.2 - -
Error (%) 0.63 5.10 2.46 3.23 0.85 1.67 3.91 0.79 1.69 3.33 - -
30SCS (s) 2.34 1.06 1.02 1.07 1.2 1.12 1.34 1.22 1.29 - - -
Test #2 Video (s) 2.32 1.06 1.01 1.1 1.23 1.11 1.33 1.18 1.27 - - -
Error (%) 0.86 0.00 0.99 2.73 2.44 0.90 0.75 3.39 1.57 - - -
7. Results
The sTUG Doctor application has been used in the Center for Aging, Huntsville, AL. The Center
led by a gerontologist provides an innovative and comprehensive model of care for adults over the
age of 65. Its primary goal is to empower the general population with the latest knowledge about
issues of aging, particularly about the use of medications and unique problems of the elderly, such as
memory disturbances, polypharmacy, falls, incontinence, dementia, depression, and mobility problems.
Specifically, sTUG Doctor is used in a pilot study aimed at fall prevention and improved mobility. It is
used to quantify the impact of medical interventions that typically include changes in medications
and community-based exercise sessions designed to improve mobility. Initial pilot study included
30 patients. Table 7 shows the results for three patients (P#1, P#2, and P#3) who participated in the
study. For each patient, we show the TUG test results taken: (a) when the patient is admitted into
the study; (b) once midway through the study; and (c) once approximately three months through the
study. The TUG results showed that all three patients significantly improved their performance in the
TUG test (e.g., almost 34% improvement of the total time d.TUG for P#1, more than 31% improvement
for P#2, and 28% improvement for P#3). Sit-to-stand (d.S2ST), stand-to–sit (d.ST2S) and total walk time
(d.WALK) durations have also been improved significantly. Moreover, total number of steps (n.STP)
taken to cover the three-meter distance has been reduced indicating the improved stability of walking.
These certainly indicate improved mobility over time.
Table 7. The TUG test results recorded during stability improvement program.
P#1 (Female, Age 79) P#2 (Female, Age 80) P#3 (Female, Age 87)
12 02 12 % 12 02 21 % 13 27 11 %
Parameter
Feb Apr May Change Feb Apr May Change Feb Mar May Change
d.TUG 16.59 13.34 10.92 ´34 15.11 11.12 10.46 ´31 20.96 16.62 15.03 ´28
d.S2ST 1.29 1.23 0.94 ´27 0.84 0.62 0.64 ´24 1.19 1.85 0.95 ´20
d.LF 0.9 0.71 0.42 ´53 0.73 0.49 0.47 ´36 0.79 1.34 0.72 ´9
d.LT 0.39 0.52 0.52 33 0.11 0.13 0.17 55 0.4 0.51 0.23 ´43
d.WK 13.79 9.88 8.76 ´36 12.59 9.91 8.06 ´36 18.29 12.02 12.53 ´31
d.ST2S 1.51 2.23 1.22 ´19 1.68 0.59 1.76 5 1.48 2.75 1.55 5
d.PS 0.32 0.58 0.39 22 0.43 0.31 0.33 ´23 0.26 0.58 0.48 85
d.SD 1.19 1.65 0.83 ´30 1.25 0.28 1.43 14 1.22 2.17 1.07 ´12
a.S2ST 61.75 42.22 42.51 ´31 47.76 49.89 51.88 9 45.37 65.8 39.39 ´13
v.LF 256.0 178.9 209.5 ´18 161.5 231.5 213.5 32 124.5 100.2 72.08 ´42
v.LT ´8.35 0 ´6.8 ´19 ´36 ´22.9 ´33.1 ´8 ´1.86 ´13.9 ´1.82 ´2
n.STP 34 25 15 ´56 35 19 13 ´63 38 19 25 ´34
n.SBT 15 12 7 ´53 17 10 7 ´59 20 10 12 ´40
The 30SCS application has been tested in our mHealth laboratory. Five healthy subjects
participated in this test. Table 8 presents the average/standard deviation of the 30SCS test cycle
parameters for those subjects. P#5 completed 16 stand ups and 15 complete cycles (i.e., from the
moment of sitting to the completion of stand-to-sit transition). The average trunk angle (a.S2ST) is also
smaller when compared to other cases. This indicates that the subject P#5 avoids leaning forward in
order to increase the number of stand ups. We observed that most subjects require more time per cycle
as the test progresses, which may be an indication of the fitness level. Figure 13 illustrates the change
of the cycle time during the 30SCS tests, as well as the fitted regression line. We estimated the slope of
the regression line by calculating ratio of the height of the line at the end to the height of the line at the
beginning. Slope values greater than one indicate that a subject needs more time to complete a test
cycle as the test progresses. In addition, an increase in the duration of individual phases (such as d.SIT,
d.S2ST, d.ST, d.ST2S, etc.) have also been noticed towards the end of the test.
Increasing Slope Regression Lines for Cycle Duration during 30SCS

2.7
RLine P#1
2.6
d.CYCL P#1
2.5 RLine P#2
d.CYCL P#2
2.4
d.CYCL [S]
RLine P#3
2.3
d.CYCL P#3
2.2
RLine P#4
2.1 d.CYCL P#4
RLine P#5
2
d.CYCL P#5
1.9
5 10 15 20 25 30
Time [S]
Figure
Figure13.
13.Regression
Regression lines
lines fitted
fitted in
in all cycle times
all cycle times (d.CYCL)
(d.CYCL)with
with95%
95%Confidence
Confidenceinterval.
interval.
Table 9 presents the 4SBT application

Table 8.results performed
The 30SCS on a healthy female 28-year-old subject
test results.
recorded for three different cases (stationary, balancing and stumbling) explained in the 4SBT
application section. TheP#1 mean
(Male, and standard
P#2 (Male,deviation
P#3 (M and SD P#4
(Male, rows in the table)
(Female, are shown to
P#5 (Female,
Parameter
illustrate variability of the displacement during 10 s of each test. We notice that the displacement
Age 32) Age 47) Age 29) Age 24) Age 28)
values inn.SUP
full tandem stand 12 (s.FTS) are lower
12 than the values 11 in a one leg13stand (s.OL). There
16 is more
variationn.CC
in the values for 12 standing when 12 a subject stumbles.
10 12
The stumbling occurs at15T = 3 in the
d.CC test. Therefore,
one-leg stand 29.02 we can say29.80
that the higher28.71
values of chest28.60
displacements28.90
are associated
d.S2STi 1.0 ˘ 0.17 1.0 ˘ 0.25 0.96 ˘ 0.2 1.0 ˘ 0.17 0.7 ˘ 0.15
with the decreased
d.LFi stability during
0.44 ˘ 0.16 the positions
0.42 ˘ 0.2 of the 4SBT
0.3 ˘ 0.16 test. 0.5 ˘ 0.15 0.31 ˘ 0.2
d.LTi 0.57 ˘ 0.04 0.6 ˘ 0.04 0.6 ˘ 0.14 0.6 ˘ 0.03 0.4 ˘ 0.02
d.ST2Si 1.2 ˘ 0.2 Table 1.1
9. The 4-stage balance
˘ 0.17 0.9 ˘ test
0.35 results. 1.4 ˘ 0.04 0.8 ˘ 0.22
d.PSi 0.45 ˘ 0.21 0.5 ˘ 0.15 0.5 ˘ 0.23 0.6 ˘ 0.03 0.41 ˘ 0.1
d.SDi Standing
0.76 Stationary
˘ 0.04 0.6Standing
˘ 0.05 with 0.5 Balance
˘ 0.14Trial Standing
0.8 ˘ 0.03 with Stumble
0.44 ˘ 0.1
d.STi s.FTS s.STS
0.29 ˘ s.TS
0.25 s.OL 0.1s.FTS
˘ 0.17 s.STS 0.4 s.TS
˘ 0.30s.OL s.FTS s.STS s.TS
0.1 ˘ 0.06 0.13 s.OL
˘ 0.1
1
d.SITi 0.02 0.12 0.05
0.30 ˘ 0.11 0.2 0.2 0.75
˘ 0.03 1.6 0.54 0.89˘ 0.8 0.1 5.6 ˘ 0.12
0.4 0.65 0.51 8.21
0.35 ˘ 0.1
2
a.S2STi 0.19 0.28 0.1
44 ˘ 5.7 0.35 1.6
48 ˘ 22.2 11.88 1.09
57.9 ˘ 10 0.21 13.78 1.17
40.3 ˘ 5.2 0.88 12.29
29.6 ˘ 3.0
3
v.LFi 0.59 0.18
186 ˘0.06
26 0.19 1720.48
˘ 47 1.94 268 9.05˘ 19 0.3 3.23
144 ˘ 182.07 8.43192 34.66
˘ 28
4
v.LTi 0.6 0.72 0.71
5.3 ˘ 5.2 1.03 12.70.21
˘ 19 8.27 2.74 3.02˘ 8.213.79 18.11 24.8
7.59 ˘ 7.4 0.731.618.93
˘ 28
5
d.CYCLi 1.63 0.41 1
2.4 ˘ 0.21 0.75 2.51.99
˘ 0.3 8.65 2.7 12.88
˘ 1.09 2.3 9.3 ˘ 0.2
2.38 12 2.67
1.9 ˘12.21
0.25
6
Slope 0.21 0.411.141.5 0.91 1.48
1.09 0.38 2.75
1.03 10.4 7.771.03 7.46 0.9 1.0512.34
7 0.67 0.02 0.21 0.86 2.99 2.14 1.94 15.82 16.22 12.28 1.75 13.84
8 0.43 0.24 1.88 2.24 0.3 0.49 1.73 3.66 16.86 10.35 3.54 11.05
Table9 9 presents
0.02 the 4SBT
0.71 2.36 application
0.73 0.76results performed
5.33 5.75 on a healthy
7.89 27.93 female
5.95 28-year-old
1.85 13.99 subject
10 0.08 0.82 0.47 2.60 5.64 3.66 8.29 12.83 14.64
recorded for three different cases (stationary, balancing and stumbling) explained 15.14 2.81 in
12.71
the 4SBT
M 0.44 0.39 0.83 1.0 1.62 4.43 4.74 6.73 13.34 16.40 2.4 15.02
application section. The mean and standard deviation (M and SD rows in the table) are shown
SD 0.45 0.26 0.79 0.77 1.57 3.75 3.87 5.84 6.83 10.4 2.22 7.02
to illustrate variability of the displacement during 10 s of each test. We notice that the displacement
values in full tandem stand (s.FTS) are lower than the values in a one leg stand (s.OL). There is more
8. Conclusions
variation in the values for standing when a subject stumbles. The stumbling occurs at T = 3 in the
Proliferation
one-leg stand test. of smartphones
Therefore, we canthat integrate
say that increasingly
the higher values ofsophisticated sensors
chest displacements arecreates new
associated
opportunities for instrumentation
with the decreased and
stability during the quantification
positions of test.
of the 4SBT standard monitoring and diagnostic
procedures. This is particularly important for the monitoring of elderly and chronic patients. In this
paper, we describe a suite of smartphone applications we developed to automate standard
recommended mobility assessment tests: TUG, 30SCS, and 4SBT. The applications provide
instantaneous feedback to the user and automatically upload the test results to the mHealth server.
The application has been tested on a group of healthy volunteers and elderly patients with recent
falls and showed promising results. By utilizing commodity smartphones, the application suite
offers an affordable tool for instantaneous quantification of the mobility assessment. Analysis of the
parameters collected over longer periods of time may help in longitudinal monitoring of mobility of
subjects and quantify the effectiveness of exercise programs and therapies.
Table 9. The 4-stage balance test results.
Standing Stationary Standing with Balance Trial Standing with Stumble

s.FTS s.STS s.TS s.OL s.FTS s.STS s.TS s.OL s.FTS s.STS s.TS s.OL
1 0.02 0.12 0.05 0.2 0.75 1.6 0.89 0.1 5.6 0.65 0.51 8.21
2 0.19 0.28 0.1 0.35 1.6 11.88 1.09 0.21 13.78 1.17 0.88 12.29
3 0.59 0.18 0.06 0.19 0.48 1.94 9.05 0.3 3.23 2.07 8.43 34.66
4 0.6 0.72 0.71 1.03 0.21 8.27 3.02 13.79 18.11 24.8 0.7 18.93
5 1.63 0.41 1 0.75 1.99 8.65 12.88 2.3 9.3 12 2.67 12.21
6 0.21 0.41 1.5 0.91 1.48 0.38 2.75 10.4 7.77 7.46 0.9 12.34
7 0.67 0.02 0.21 0.86 2.99 2.14 1.94 15.82 16.22 12.28 1.75 13.84
8 0.43 0.24 1.88 2.24 0.3 0.49 1.73 3.66 16.86 10.35 3.54 11.05
9 0.02 0.71 2.36 0.73 0.76 5.33 5.75 7.89 27.93 5.95 1.85 13.99
10 0.08 0.82 0.47 2.60 5.64 3.66 8.29 12.83 14.64 15.14 2.81 12.71
M 0.44 0.39 0.83 1.0 1.62 4.43 4.74 6.73 13.34 16.40 2.4 15.02
SD 0.45 0.26 0.79 0.77 1.57 3.75 3.87 5.84 6.83 10.4 2.22 7.02
8. Conclusions
Proliferation of smartphones that integrate increasingly sophisticated sensors creates new
opportunities for instrumentation and quantification of standard monitoring and diagnostic procedures.
This is particularly important for the monitoring of elderly and chronic patients. In this paper,
we describe a suite of smartphone applications we developed to automate standard recommended
mobility assessment tests: TUG, 30SCS, and 4SBT. The applications provide instantaneous feedback to
the user and automatically upload the test results to the mHealth server. The application has been
tested on a group of healthy volunteers and elderly patients with recent falls and showed promising
results. By utilizing commodity smartphones, the application suite offers an affordable tool for
instantaneous quantification of the mobility assessment. Analysis of the parameters collected over
longer periods of time may help in longitudinal monitoring of mobility of subjects and quantify the
effectiveness of exercise programs and therapies.
Acknowledgments: This work was supported in part by National Science Foundation grants CNS-1205439
and CNS-1217470. Authors would like to thank Mladen Milosevic for developing a computing infrastructure
which included an mHealth server and a number of mHealth applications for mobile health and wellness
monitoring, including the original sTUG smartphone application. We thank Zaheer Khan from Center for Elder
Care, Huntsville, AL and Karen Frith and Amy Hunter for providing TUG records from longitudinal study of
assessment of mobility used as test cases in this paper.
Author Contributions: Priyanka Madhushri developed the 30SCS and 4SBT smartphone applications and worked
on paper preparation. Armen Dzhagaryan proposed the test parameters for 30SCS application, developed signal
processing algorithms for 30SCS, and helped in preparing figures. Emil Jovanov and Aleksandar Milenkovic
proposed the idea, established the mHealth framework, and provided guidance and management throughout the
completion of the project. All authors have read and approved the final manuscript.
Conflicts of Interest: The authors declare no conflict of interest.
References
1. Gibson, M.J.; Andres, R.O.; Isaacs, B.; Radebaugh, T.; Wormpetersen, J. The prevention of falls in later life.
A report of the Kellogg International Work Group on the Prevention of Falls by the Elderly. Dan. Med. Bull.
1987, 34 (Suppl. 4), 1–24.
2. Hsu, C.L.; Nagamatsu, L.S.; Davis, J.C.; Liu-Ambrose, T. Examining the relationship between specific
cognitive processes and falls risk in older adults: A systematic review. Osteoporos. Int. J. Establ. Result Coop.
Eur. Found. Osteoporos. Natl. Osteoporos. Found. USA 2012, 23, 2409–2424. [CrossRef] [PubMed]
3. Hamacher, D.; Singh, N.B.; Van Dieën, J.H.; Heller, M.O.; Taylor, W.R. Kinematic measures for assessing gait
stability in elderly individuals: A systematic review. J. R. Soc. Interface R. Soc. 2011, 8, 1682–1698. [CrossRef]
[PubMed]
4. Vincent, G.K.; Velkoff, V.A. The Next Four Decades: The Older Population in the United States: 2010 to 2050;
U.S. Department of Commerce. Economics and Statistics Administration, U.S. Census Bureau: Suitland, MD,
USA, 2010.
5. Alexander, B.H.; Rivara, F.P.; Wolf, M.E. The cost and frequency of hospitalization for fall-related injuries in
older adults. Am. J. Public Health 1992, 82, 1020–1023. [CrossRef] [PubMed]
6. Habib, M.A.; Mohktar, M.S.; Kamaruzzaman, S.B.; Lim, K.S.; Pin, T.M.; Ibrahim, F. Smartphone-Based
Solutions for Fall Detection and Prevention: Challenges and Open Issues. Sensors 2014, 14, 7181–7208.
[CrossRef] [PubMed]
7. Schneider, D.L. Preventing Falls: Time for Action. Available online: http://www.todaysgeriatricmedicine.
com/archive/110113p34.shtml (accessed on 31 January 2016).
8. Thrane, G.; Joakimsen, R.M.; Thornquist, E. The association between timed up and go test and history of
falls: The Tromsø study. BMC Geriatr. 2007, 7, 1. [CrossRef] [PubMed]
9. Millor, N.; Lecumberri, P.; Gómez, M.; Martínez-Ramírez, A.; Izquierdo, M. An evaluation of the 30-s
chair stand test in older adults: Frailty detection based on kinematic parameters from a single inertial unit.
J. NeuroEng. Rehabil. 2013, 10, 86. [CrossRef] [PubMed]
10. CDC The 4-Stage Balance Test. Available online: http://www.cdc.gov/steadi/pdf/4-stage_balance_test-a.
pdf (accessed on 31 January 2016).
11. Zampieri, C.; Salarian, A.; Carlson-Kuhta, P.; Nutt, J.G.; Horak, F.B. Assessing mobility at home in people
with early Parkinson’s disease using an instrumented Timed Up and Go test. Parkinsonism Relat. Disord.
2011, 17, 277–280. [CrossRef] [PubMed]
12. Milenković, A.; Otto, C.; Jovanov, E. Wireless sensor networks for personal health monitoring: Issues and an
implementation. Comput. Commun. 2006, 29, 2521–2533. [CrossRef]
13. Jovanov, E.; Poon, C.; Yang, G.-Z.; Zhang, Y.T. Guest Editorial Body Sensor Networks: From Theory to
Emerging Applications. IEEE Trans. Inf. Technol. Biomed. 2009, 13, 859–863. [CrossRef] [PubMed]
14. Jovanov, E.; Milenkovic, A. Body area networks for ubiquitous healthcare applications: Opportunities and
Challenges. J. Med. Syst. 2011, 35, 1245–1254. [CrossRef] [PubMed]
15. Bosch, S.; Marin-Perianu, M.; Marin-Perianu, R.; Havinga, P.; Hermens, H. Keep on moving! Activity
monitoring and Stimulation using wireless sensor networks. In Smart Sensing and Context, Proceedings of
the 4th European Conference on Smart Sensing and Context (EuroSSC’09), Guildford, UK, 16–18 September
2009; Lecture Notes in Computer Science; Springer-Verlag: Berlin, Germany; Heidelberg, Germany, 2009;
Volume 5741, pp. 11–23.
16. Pitta, F.; Troosters, T.; Probst, V.S.; Spruit, M.A.; Decramer, M.; Gosselink, R. Quantifying physical activity in
daily life with questionnaires and motion sensors in COPD. Eur. Respir. J. 2006, 27, 1040–1055. [CrossRef]
[PubMed]
17. Madhushri, P.; Ahmed, B.; Penzel, T.; Jovanov, E. Periodic leg movement (PLM) monitoring using a
distributed body sensor network. Conf. Proc. IEEE Eng. Med. Biol. Soc. (EMBC) 2015, 2015, 1837–1840.
18. Al-Jawad, A.; Adame, M.R.; Romanovas, M.; Hobert, M.; Maetzler, W.; Traechtler, M.; Moeller, K.;
Manoli, Y. Using multi-dimensional dynamic time warping for TUG test instrumentation with inertial
sensors. In Proceedings of the 2012 IEEE Conference on Multisensor Fusion and Integration for Intelligent
Systems (MFI), Hamburg, Germany, 13–15 September 2012; pp. 212–218.
19. Zijlstra, A.; Mancini, M.; Lindemann, U.; Chiari, L.; Zijlstra, W. Sit-stand and stand-sit transitions in older
adults and patients with Parkinson’s disease: Event detection based on motion sensors versus force plates.
J. NeuroEng. Rehabil. 2012, 9, 75. [CrossRef] [PubMed]
20. Tacconi, C.; Mellone, S.; Chiari, L. Smartphone-Based Applications for Investigating Falls and Mobility.
In Proceedings of the 2011 5th International Conference on Pervasive Computing Technologies for Healthcare
(PervasiveHealth) and Workshops, Dublin, Ireland, 23–26 May 2011.
21. Wongpatikaseree, K.; Lim, A.O.; Tan, Y.; Kanai, H. Range-based algorithm for posture classification and
fall-down detection in smart homecare system. In Proceedings of the 1st IEEE Global Conference on
Consumer Electronics, Tokyo, Japan, 2–5 October 2012; pp. 243–247.
22. Jovanov, E.; Milenkovic, A.; Otto, C.; de Groen, P.C. A wireless body area network of intelligent motion
sensors for computer assisted physical rehabilitation. J. NeuroEng. Rehabil. 2005, 2, 6. [CrossRef] [PubMed]
23. Santhiranayagam, B.K.; Lai, D.T.H.; Jiang, C.; Shilton, A.; Begg, R. Automatic detection of different walking
conditions using inertial sensor data. In Proceedings of the 2012 International Joint Conference on Neural
Networks (IJCNN), Brisbane, Australia, 10–15 June 2012; pp. 1–6.
24. Weiss, A.; Herman, T.; Plotnik, M.; Brozgol, M.; Maidan, I.; Giladi, N.; Gurevich, T.; Hausdorff, J.M. Can an
accelerometer enhance the utility of the Timed up & Go Test when evaluating patients with Parkinson’s
disease? Med. Eng. Phys. 2010, 32, 119–125. [PubMed]
25. Zampieri, C.; Salarian, A.; Carlson-Kuhta, P.; Aminian, K.; Nutt, J.G.; Horak, F.B. The instrumented timed up
and go test: Potential outcome measure for disease modifying therapies in Parkinson’s disease. J. Neurol.
Neurosurg. Psychiatry 2010, 81, 171–176. [CrossRef] [PubMed]
26. Marschollek, M.; Nemitz, G.; Gietzelt, M.; Wolf, K.H.; Meyer Zu Schwabedissen, H.; Haux, R. Predicting
in-patient falls in a geriatric clinic: A clinical study combining assessment data and simple sensory gait
measurements. Z. Gerontol. Geriatr. 2009, 42, 317–321. [CrossRef] [PubMed]
27. Bao, L.; Intille, S.S. Activity Recognition from User-Annotated Acceleration Data. In Pervasive Computing;
Ferscha, A., Mattern, F., Eds.; Lecture Notes in Computer Science; Springer: Berlin, Germany; Heidelberg,
Germany, 2004.
28. Ghasemzadeh, H.; Loseu, V.; Jafari, R. Structural action recognition in body sensor networks: Distributed
classification based on string matching. IEEE Trans. Inf. Technol. Biomed. 2010, 14, 425–435. [CrossRef]
[PubMed]
29. Lester, J.; Choudhury, T.; Borriello, G. A practical approach to recognizing physical activities. In Pervasive
Computing, Proceedings of the 4th International Conference on Pervasive Computing (PERVASIVE’06),
Dublin, Ireland, 7–10 May 2006; Lecture Notes in Computer Science; Springer: Berlin, Germany; Heidelberg,
Germany, 2006; Volume 3968, pp. 1–16.
30. Aminian, K.; Najafi, B. Body Movement Monitoring System and Method. U.S. Patent 7141026 B2,
1 October 2001.
31. Florentino-Liano, B.; O’Mahony, N.; Artes-Rodriguez, A. Human activity recognition using inertial sensors
with invariance to sensor orientation. In Proceedings of the 2012 3rd International Workshop on Cognitive
Information Processing (CIP), Baiona, Spain, 28–30 May 2012.
32. Fuentes, D.; Gonzalez-Abril, L.; Angulo, C.; Ortega, J.A. Online motion recognition using an accelerometer
in a mobile device. Expert Syst. Appl. 2012, 39, 2461–2465. [CrossRef]
33. Godfrey, A.; Bourke, A.K.; Ólaighin, G.M.; van de Ven, P.; Nelson, J. Activity classification using a single
chest mounted tri-axial accelerometer. Med. Eng. Phys. 2011, 33, 1127–1135. [CrossRef] [PubMed]
34. Milosevic, M.; Milenkovic, A.; Jovanov, E. mHealth @ UAH: Computing Infrastructure for Mobile Health
and Wellness Monitoring. XRDS 2013, 20, 43–49. [CrossRef]
35. Podsiadlo, D.; Richardson, S. The timed “Up & Go”: A test of basic functional mobility for frail elderly
persons. J. Am. Geriatr. Soc. 1991, 39, 142–148. [PubMed]
36. CDC The Timed Up and Go (TUG) Test. Available online: http://www.cdc.gov/steadi/pdf/tug_test-a.pdf
(accessed on 31 January 2016).
37. Milosevic, M.; Jovanov, E.; Milenkovic, A. Quantifying Timed-Up-and-Go test: A smartphone
implementation. In Proceedings of the 2013 IEEE International Conference on Body Sensor Networks
(BSN), Cambridge, MA, USA, 6–9 May 2013.
38. Csuka, M.; McCarty, D.J. Simple method for measurement of lower extremity muscle strength. Am. J. Med.
1985, 78, 77–81. [CrossRef]
39. Lord, S.R.; Murray, S.M.; Chapman, K.; Munro, B.; Tiedemann, A. Sit-to-Stand Performance Depends on
Sensation, Speed, Balance, and Psychological Status in Addition to Strength in Older People. J. Gerontol. A
Biol. Sci. Med. Sci. 2002, 57, M539–M543. [CrossRef] [PubMed]
40. Whitney, S.L.; Wrisley, D.M.; Marchetti, G.F.; Gee, M.A.; Redfern, M.S.; Furman, J.M. Clinical measurement
of sit-to-stand performance in people with balance disorders: Validity of data for the Five-Times-Sit-to-Stand
Test. Phys. Ther. 2005, 85, 1034–1045. [PubMed]
41. Buatois, S.; Miljkovic, D.; Manckoundia, P.; Gueguen, R.; Miget, P.; Vançon, G.; Perrin, P.; Benetos, A.
Five times sit to stand test is a predictor of recurrent falls in healthy community-living subjects aged 65 and
older. J. Am. Geriatr. Soc. 2008, 56, 1575–1577. [CrossRef] [PubMed]
42. CDC 30-Second Chair Stand Test. Available online: http://www.cdc.gov/steadi/pdf/30_second_chair_
stand_test-a.pdf (accessed on 31 January 2016).
© 2016 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC-BY) license (http://creativecommons.org/licenses/by/4.0/).
View publication stats

Information 07 00047

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Information 07 00047

Uploaded by

Copyright:

Available Formats

See discussions, stats, and author profiles for this publication at: https://www.researchgate.

An mHealth Tool Suite for Mobility Assessment

Article in Information (Switzerland) · July 2016

Priyanka Madhushri Armen A Dzhagaryan

SEE PROFILE SEE PROFILE

Emil Jovanov Aleksandar Milenkovic

SEE PROFILE SEE PROFILE

The user has requested enhancement of the downloaded file.

Information 2016, 7, 47; doi:10.3390/info7030047 www.mdpi.com/journal/information

(a) (b) (c)

(d) (e) (f)

Figure 3. Timed Up and Go (TUG) test phases.

(a) (b) (c)

Table 1. Parameters for TUG characterization.

Figure 6. The 30-Second Chair Stand (30SCS) test phases.

(a) (b) (c)

Parameter Description Units

5. The 4-Stage Balance Test (4SBT) Application

Figure 9. The 4-Stage Balance Test (4SBT) feet positions.

(a) (b) (c)

Table 4. Parameters for 4-Stage Balance Test characterization.

Parameter Description Units

(a) (b) (c)

6. Application Suite Verification and Validation

6. Application Suite Verification and Validation

- - Test#1 Test#2 Test#3

Table 6. Testing the 30SCS application: sit-to-stand transition times (d.S2STi).

S#1 (Female, 28) - #1 #2 #3 #4 #5 #6 #7 #8 #9 #10 #11 #12

Increasing Slope Regression Lines for Cycle Duration during 30SCS

2.5 RLine P#2

2.1 d.CYCL P#4

Table 9 presents the 4SBT application

Table 9. The 4-stage balance test results.

Standing Stationary Standing with Balance Trial Standing with Stumble

View publication stats

You might also like