sensors
Article
A Systematic Approach to the Design and
Characterization of a Smart Insole for Detecting
Vertical Ground Reaction Force (vGRF) in
Gait Analysis
Anas M. Tahir 1 , Muhammad E. H. Chowdhury 1 , Amith Khandakar 1 , Sara Al-Hamouz 1 ,
Merna Abdalla 1 , Sara Awadallah 1 , Mamun Bin Ibne Reaz 2 and Nasser Al-Emadi 1, *
1
2
*
Department of Electrical Engineering, Qatar University, Doha 2713, Qatar; a.tahir@qu.edu.qa (A.M.T.);
mchowdhury@qu.edu.qa (M.E.H.C.); amitk@qu.edu.qa (A.K.); sa1507714@student.qu.edu.qa (S.A.-H.);
ma1508307@student.qu.edu.qa (M.A.); sa1509524@student.qu.edu.qa (S.A.)
Department of Electrical, Electronic & Systems Engineering, Universiti Kebangsaan Malaysia, Bangi,
Selangor 43600, Malaysia; mamun@ukm.edu.my
Correspondence: alemadin@qu.edu.qa; Tel.: +974-4403-4213
Received: 7 December 2019; Accepted: 28 January 2020; Published: 11 February 2020
Abstract: Gait analysis is a systematic study of human locomotion, which can be utilized in various
applications, such as rehabilitation, clinical diagnostics and sports activities. The various limitations
such as cost, non-portability, long setup time, post-processing time etc., of the current gait analysis
techniques have made them unfeasible for individual use. This led to an increase in research interest in
developing smart insoles where wearable sensors can be employed to detect vertical ground reaction
forces (vGRF) and other gait variables. Smart insoles are flexible, portable and comfortable for gait
analysis, and can monitor plantar pressure frequently through embedded sensors that convert the
applied pressure to an electrical signal that can be displayed and analyzed further. Several research
teams are still working to improve the insoles’ features such as size, sensitivity of insoles sensors,
durability, and the intelligence of insoles to monitor and control subjects’ gait by detecting various
complications providing recommendation to enhance walking performance. Even though systematic
sensor calibration approaches have been followed by different teams to calibrate insoles’ sensor,
expensive calibration devices were used for calibration such as universal testing machines or infrared
motion capture cameras equipped in motion analysis labs. This paper provides a systematic design
and characterization procedure for three different pressure sensors: force-sensitive resistors (FSRs),
ceramic piezoelectric sensors, and flexible piezoelectric sensors that can be used for detecting vGRF
using a smart insole. A simple calibration method based on a load cell is presented as an alternative
to the expensive calibration techniques. In addition, to evaluate the performance of the different
sensors as a component for the smart insole, the acquired vGRF from different insoles were used to
compare them. The results showed that the FSR is the most effective sensor among the three sensors
for smart insole applications, whereas the piezoelectric sensors can be utilized in detecting the start
and end of the gait cycle. This study will be useful for any research group in replicating the design of
a customized smart insole for gait analysis.
Keywords: gait analysis; characterization; smart insole; vertical ground reaction forces; force sensitive
resistors; piezoelectric sensors; sensor calibration
Sensors 2020, 20, 957; doi:10.3390/s20040957
www.mdpi.com/journal/sensors
Sensors 2020, 20, 957
2 of 25
1. Introduction
Gait analysis offers an opportunity for assessment of the act of walking, one of the most important
features of the individual’s use pattern that displays posture in action. By identifying gait kinetics, gait
kinematics and musculoskeletal activity, gait analysis can be utilized in various applications, such as
rehabilitation, clinical diagnostics and sport activities [1]. Gait kinetics studies the forces and moments
that results in movement of lower extremities during gait cycle. Vertical ground reaction forces
(vGRFs) are the forces between the foot and ground which can be obtained by wearable sensors [2]
and are considered as the main measurement in kinetic analysis. Gait kinetics have recently become
a convenient tool for biomedical research and clinical practice. Different research teams studied the
ability to diagnose or early detection of various diseases using gait analysis [3–5]. Some research teams
used gait analysis in fall detection of elderly people, one of the most common domestic accidents
among the elderly. With smart insoles, the fall event can be detected and doctors or personal who takes
care of the elderly can be notified to take action. In athletic sports where walking, running, jumping
and throwing are involved, gait analysis can be utilized to recognize an athlete’s faulty movement and,
accordingly, enhance it. In addition, gait analysis can play positive role in the rehabilitation process for
several diseases and complications.
Recently, with the development in sensor technologies, gait analysis using wearable systems
became an effective approach [6–8]. Various types of wearable sensors such as force sensors, strain
gauges, magneto-resistive sensors, accelerometers, gyroscopes, inclinometers etc. can analyze different
gait characteristics. Accelerometers were used to conduct gait analysis studies, in which they were
attached to feet or legs to measure the acceleration or velocity of human lateral movements during gait
cycles [4]. Gyroscopes were used in gait analysis to measure the changes in orientation of lower body
extremes with respect to the vertical axis. Goniometers measured the relative rotational motion between
different body segments [2]. Electromagnetic tracking systems were developed as 3D measurement
device that can be applied in the kinematic study of body movements [9].
Gait analysis is typically carried out using a force plate system or multi-camera-based system
to capture the ground reaction forces (GRF) during different gait cycles. However, this method
requires a costly set up and long post-processing time and can measure only limited number of strides.
Therefore, it is not affordable by individuals for personal use [3,8,10]. Instrumented trade mills with
few force plates laid on the trade mill are used by different research groups to mitigate the limitations
of conventional force plates [2], but with treadmills restrictions are still present as subjects need to walk
in a straight line where direction changes and turning cannot be realized. This led to an increase in
research interest towards developing smart insoles, where wearable sensors can be employed to detect
vGRF, joint movements, acceleration of lower extremities, and other gait variables [3,4,11,12]. vGRF is a
useful tool to assess the health conditions of the patient, to enhance the performance of athletes [13–15].
Among different solutions for vGRF measurement, smart insoles have several extra advantages over
force plates and multi-camera systems. Although force plates can measure shear forces and pressure
changes, smart insoles are portable and capable of tracking motions and measuring pressure without
rigid mounting, whereas the camera-based system requires large space for set-up along with long
post-processing time. The smart insole offers flexible, portable, and comfortable solution for vGRF
measurement. It is designed to monitor, process and display plantar pressure using pressure sensors
embedded in the insole [3,4,11,12]. Recently, several off-the-shelf smart insoles have been offered by
some companies (e.g., F-scan [16], MoveSole [17], Bonbouton [18], FeetMe [19] etc.), however, the
commercial systems are very expensive for individual use, making it difficult for a home setting.
The aim of this study is to design and characterize smart insoles to detect vGRF during gait,
with three different types of low-cost commercial force sensor: force-sensitive resistors (FSRs) [20],
ceramic piezoelectric sensors [21], and flexible piezoelectric sensors [22]. All three types of sensor were
calibrated before checking their suitability for smart insole application. A simple low-cost calibration
method based on load cells is presented, mitigating the need to use expensive calibration devices or
Motion Analysis Labs as a calibration reference. This work provides a systematic approach for sensor
Sensors 2020, 20, 957
3 of 25
calibration guides, which can be replicated easily by other researchers to perform studies on smart
insoles or other body-sensing technologies. To the best of our knowledge, this is the first article to
compare three different low-cost commercially available force sensors for smart insole application.
The remainder of the article is organized into five sections. In Section 2, a comprehensive review
of the recent works with smart insoles to detect vGRF in gait cycles are summarized. In Section 3, the
experimental details for sensors calibration and insole characterization are presented. In Section 4, the
mathematical analysis of each insole characterization and sensor calibration are explained. Results and
a discussion are presented in Section 5. Finally, we conclude with future recommendations in Section 6.
2. Literature Review
Several research teams focused on fabricating and synthesizing the sensing parts or sensing
fabrics of the smart insoles [23–25]. Sensing fabrics are fibers/yarns with sensing technologies or
electrical components made of fabric materials, offering a flexible alternative to comfortably measuring
human movement. Usually, piezoelectric, piezoresistive and piezo-capacitive materials are used to
fabricate the sensing parts of the sensing fabrics, due to their elastic properties [26,27]. Shu et al. [26]
implemented a low-cost insole with high pressure sensitivity using a fabric pressure sensing array
made by the researchers with a pressure range of 10 Pa to 1000 kPa. It is attached to six locations
corresponding to a polyimide film circuit board that takes the shape of the foot. They were able to
measure the peak pressure, mean pressure, center of pressure (COP), and illustrate different pressure
levels occurring at the six-targeted areas. However, the quality of the gait cycle records was poor,
with irregular peak values, where the common gait shape with two peaks of the heel strike and toe
off cannot be distinguished. Kessler et al. [27] demonstrated a low-cost flexible insole, made with
Velostat and conductive ink electrodes printed on polyethylene terephthalate (PET) substrate. However,
repeatability was a major problem and they proposed an averaging method to reduce the repeatability
issue. However, the proposed method does not provide a generic solution for the force-sensing
problem, it can be utilized only with periodic forces where spatial information is the key. On the other
hand, some research teams used low-cost flexible force sensors to design the smart insoles [28–30]
using commercially available piezoresistive [20], piezoelectric [21,22], capacitive transducers [2], fiber
brag grating [5,31] sensors.
Piezoelectric force sensors are materials that generate electric charges when stressed. However,
there are a few factors which limit the usage of piezoelectric sensors in smart insoles. The parasitic
effect of piezo materials neutralizes the generated charge within a short time. Therefore, sophisticated
electronics are needed to extract resultant charges, and this makes it difficult to use these sensors in
measuring static or slow varying forces. In addition, protection circuits are needed, since piezo sensors
generate high voltage values, which might reach above 100 V with peak vGRF values. Capacitive force
sensors are another alternative force sensor, consisting of parallel capacitor plates that changes the
capacitance in correspondence to applied force/weight. However, they need complex conditioning
circuits and are highly subject to noise [20].
A commonly used body-sensing technology is the piezo-resistive sensor or FSR, which changes its
conductivity based on the applied force. FSR is a polymer thick film (PTF) that is used to measure the
applied force in different applications such as human touch and medical applications, industrial and
robotics applications, and automotive electronics. The main advantages of FSRs are: thin size, very
good shock resistance, low power requirement, fast response to force changes, robustness against noise,
simple conditioning circuits, ability to fabricate using flexible materials, and low unit cost compared to
other commercial force sensors [20]. However, these sensors have some disadvantages that need to be
compensated for, such as non-linear behavior and repeatability error [3].
Bamberg et al. [4] used a combination of different FSRs, piezo electric sensors, accelerometers
and gyroscopes to determine the vGRF. The main advantage of this approach is that it enables the
detection of heel strike and toe off events in each gait cycle. In addition, it helps in estimating foot
orientation and position. Even though gait variability can be analyzed by walking in a straight line,
Sensors 2020, 20, 957
4 of 25
gait analysis concentrating merely on straight walking or running may not be adequate to interpret gait
variability, since changing walking directions or turning have effects on extrinsic gait variability [11].
Similar research was done recently in [32], where the research group used the FSR sensor to develop
the smart foot sole which transmits wirelessly the vGRF to a computer, and the patients were asked to
walk on treadmill during the signal acquisition. Liu et al. [11] developed a wearable measuring insole
using five triaxial force sensors in each shoe capable of measuring GRF and center of pressure (COP)
on insole. The GRF results showed a great correspondence between the insole and the reference data.
Kim et al. [33] conducted a similar study, where they have used similar triaxial force sensors and the
sensors performance were tested on seven healthy male subjects. An in-shoe plantar measurement
sensor with 64 sensing points made from an optoelectronics transducer covered with silicon in a matrix
form covering 80% of contact region between the foot and the insole and handling capability of 1MPa
was implemented by De Rossi et al. to measure COP and vGRF [5]. Howel et al. [3] demonstrated the
design of a wearable smart insole using low-cost FSRs for gait analysis. This provided subject-specific
linear regression models to determine the vGRF accurately using simultaneous collected data from
motion analysis laboratory. However, insufficient information was given about the sensors calibration
and the hardware design of the insole and the wireless system to transmit the data to host PC, making
it difficult for other researchers to replicate the work.
Even though systematic sensor calibration with clear steps was followed by different research
teams, expensive calibration devices were used to calibrate the force sensors. Some research teams
carried out the experiments on the smart insoles in motion analysis labs, where simultaneous data
collection from infrared motion capture cameras/RGB depth camera and force plates were done as
reference measurement for the collected insole data [34,35]. In addition, some research teams used a
universal testing machine to apply incremental weight values to sensor active area during calibration.
Barnea et al. [36] used the CETR Universal Micro-Tribometer (UMT)-2 micro tribometer) device for
calibrations, that can apply precise weights in X, Y and Z directions. Marco et al. [5] performed the
sensor calibrations using robotic platform that can precisely apply controllable loads to the desired
positions. Parmar et al. [37] evaluated the performance of 5 different commercial FSRs during static
and dynamic loading with reliable test setups that can mimic realistic conditions when applying
pressure on human limbs. The sensors were evaluated quantitatively based on their accuracy, drift,
and repeatability behaviors. The tested sensors showed lower accuracy levels with static pressures
compared to the dynamic pressure test, with high drift values. This necessitates the need for further
study and analysis on the use of FSRs for static pressure applications.
3. Methodology
This section demonstrates the design of a complete system describing the main blocks of the smart
insole along with illustrations of sensor calibration and insole characterization process.
3.1. Smart Insole Sub-System
Figure 1 shows the complete block diagram of the system, where the pressure sensor array
was placed in a customized shoe above the control circuit. Pressure data were digitized through a
microcontroller before they were sent wirelessly to a host computer for post processing and analysis.
This subsystem was powered by a battery with the help of a power management unit. Pressure data
were analyzed to extract various gait characteristics for different gait applications.
Sensors 2020, 20, 957
Sensors 2020, 20, 957
5 of 25
5 of 25
relation between the applied force and the sensor’s resistance. In addition, no direct relationship is
relation between the applied force and the sensor’s resistance. In addition, no direct relationship is
5 of 25
sensor’s datasheet. Therefore, proper calibration must be done prior to the sensor
provided in the sensor’s datasheet. Therefore, proper calibration must be done prior to the sensor
usage.
usage.
Sensors
2020, 20,
provided
in 957
the
Figure1.1.Smart
Smartinsole
insoleblock
blockdiagram.
diagram.
Figure
Figure 1. Smart insole block diagram.
3.1.1. B.
Pressure-Sensing
Array Sensor
Ceramic Piezoelectric
B. Ceramic Piezoelectric Sensor
A piezoelectric
element
is a sensor
that
produces
an
alternating
voltage in response to an applied
The
vGRF during
gait cycles
can bethat
sensed
usingan
one
of three alternatives:
A piezoelectric
element
is a sensor
produces
alternating
voltage in response to an applied
dynamic
pressure or vibration.
With
applications related to dynamic forces, the piezoelectric sensor
A. Force-Sensitive
Resistor
(FSR)
dynamic
pressure or vibration.
With
applications related to dynamic forces, the piezoelectric sensor
is highly
recommended.
When
a
force
applied
to
the piezoelectric
crystal
element,
the net movement
The FSR
exhibits a decrease
resistance
asto
the
force tocrystal
the surface
of the
increases.
is highly
recommended.
When ain
force
applied
theapplied
piezoelectric
element,
thesensor
net movement
of
both
positive
and
negative
ions
occurs.
When
there
is
a
constant
or
zero
pressure,
the dipole
is not
FSR
sensors
from
Interlink
Electronics
[20]
were
used
in
this
study
as
shown
in
Figure
2A.
The
sensors
of both positive and negative ions occurs. When there is a constant or zero pressure, the dipole
is not
formed
[38]. Itround
is important
to
mention
that the
force
plate
is originally
made of piezoelectric
material
have
a flexible
activeto
area
of diameter
12.7
mm
to detect
the applied
with a twomaterial
flexible
formed
[38]. It is
important
mention
that the
force
plate
is originally
madeforce,
of piezoelectric
mounted
between
two
metal
plates
to
produce
three‐dimensional
forces
with
a
special
mechanical
lead
wires
to
connect
the
sensor
to
the
acquisition
circuit.
A
FSR
exhibits
a
non-linear
relation
between
mounted between two metal plates to produce three‐dimensional forces with a special mechanical
arrangement
[39]. This
comes
in different
sizes;
however,
ceramic
piezoelectriciselement
with
12.8
the
applied force
sensor’s
resistance.
Inhowever,
addition,aano
direct piezoelectric
relationship
provided
in12.8
the
arrangement
[39].and
Thisthe
comes
in different
sizes;
ceramic
element
with
mm
electrode
diameter
would
be
suitable
to
obtain
a
high‐resolution
pressure
map
as
shown
in
sensor’s
datasheet.
Therefore,
calibration
must abehigh‐resolution
done prior to the
sensor usage.
mm electrode
diameter
wouldproper
be suitable
to obtain
pressure
map as shown in
Figure
2b.
B. 2b.Ceramic Piezoelectric Sensor
Figure
C. Micro‐Electromechanical
Systems
(MEMS) Sensor
A
element is a sensor
that (MEMS)
produces
an alternating voltage in response to an applied
C. piezoelectric
Micro‐Electromechanical
Systems
Sensor
The
micro‐electromechanical
systems
(MEMS)
sensor
is a newforces,
member
of piezoelectric
sensors
dynamic
pressure
or
vibration.
With
applications
related
to dynamic
the
is
The micro‐electromechanical systems (MEMS) sensor
is a new member
ofpiezoelectric
piezoelectricsensor
sensors
family
(Figure
2c).
Similar
to
ceramic
piezo
electric
sensors,
it
converts
mechanical
forces
into
highly
recommended.
When
a
force
applied
to
the
piezoelectric
crystal
element,
the
net
movement
of
family (Figure 2c). Similar to ceramic piezo electric sensors, it converts mechanical forces into
electrical
signals.
However,
the occurs.
MEMS sensor
can
detect
forces in x, zero
y or z axes generating
electrical
both
positive
andHowever,
negative ions
Whencan
there
is a constant
the dipole
is not
electrical
signals.
the MEMS
sensor
detect
forces in or
x, y or zpressure,
axes generating
electrical
impulses
with
positive
or
negative
amplitudes
depending
on
the
force
direction
on
a
certain
axis[40].
formed
[38].
It
is
important
to
mention
that
the
force
plate
is
originally
made
of
piezoelectric
material
impulses with positive or negative amplitudes depending on the force direction on a certain axis[40].
MEMS sensors
are useful
for plates
detecting
human motion
sensor due to theirwith
flexibility,
wide frequency
mounted
between
metal
to human
produce
three-dimensional
a special
MEMS sensors
are two
useful
for detecting
motion
sensor due toforces
their flexibility,
widemechanical
frequency
range
(0.001
Hz
to
10
MHz),
low
acoustic
impedance,
high
mechanical
strengths,
and
high12.8
stability
arrangement
[39].
This
comes
in
different
sizes;
however,
a
ceramic
piezoelectric
element
with
mm
range (0.001 Hz to 10 MHz), low acoustic impedance, high mechanical strengths, and high stability
resisting
moisture,
etc.
[40].
electrode
diameter
would
be
suitable
to
obtain
a
high-resolution
pressure
map
as
shown
in
Figure
2B.
resisting moisture, etc. [40].
A
A
B
B
C
C
Figure 2. (A) Force‐sensitive resistor (FSR) sensor from Interlink Electronics [20], (B) piezo‐electric
piezo‐electric
Figure 2. (A) Force‐sensitive
Force-sensitive resistor
resistor (FSR)
(FSR) sensor
sensor from
from Interlink
Interlink Electronics
Electronics [20]
[20],, (B) piezo-electric
sensor from Murata Manufacturing Co.[38], (C) micro‐electromechanical systems (MEMS) sensor
sensor from Murata
[38],, (C) micro-electromechanical
Murata Manufacturing
Manufacturing Co. [38]
micro‐electromechanical systems (MEMS) sensor
LDT0‐028K from Measurement Specialties Inc. [40].
LDT0-028K
LDT0‐028K from Measurement Specialties
Specialties Inc.
Inc. [40].
[40].
C.
Micro-Electromechanical Systems (MEMS) Sensor
Sensors 2020, 20, 957
6 of 25
The micro-electromechanical systems (MEMS) sensor is a new member of piezoelectric sensors
family (Figure 2C). Similar to ceramic piezo electric sensors, it converts mechanical forces into electrical
signals. However, the MEMS sensor can detect forces in x, y or z axes generating electrical impulses
with positive or negative amplitudes depending on the force direction on a certain axis [40]. MEMS
sensors are useful for detecting human motion sensor due to their flexibility, wide frequency range
(0.001 Hz to 10 MHz), low acoustic impedance, high mechanical strengths, and high stability resisting
moisture, etc. [40].
3.1.2. Data Acquisition System
A. Microcontroller (MCU):
A microcontroller (MCU) was used to collect the data from the sensor and to send to the computer
for classification. Simblee is a very compact and powerful ARM Cortex-M0 MCU with a six channels
10-bit analog-to-digital converter (ADC). It is featured with an inter-integrated circuit (I2 C) and serial
peripheral interface (SPI) communication interface, which were required for 9-degree of freedom (DOF)
module. Moreover, it has an incorporated Bluetooth low energy (BLE) 4.0 module, which can be
utilized to send data to the computer. This MCU operates on a power supply between +2.1 to 3.6 V.
B. Multiplexer (MUX)
Since MCU has a limited number of ADC channels whereas the number of sensors is needed for
better spatial resolution of smart insole, it is suggested to use multiplexers (MUX) to reduce the number
of required channels in MCU. A MUX allows several inputs in parallel to be routed into a single output
depending on the input combinations of the data selectors. Active area of these sensors are close
and sixteen sensors were used to create sensors’ array for each leg insole to obtain a high-resolution
pressure map. Therefore, the CD74HC4067 multiplexer from Texas Instruments with 16 input channels
was used in this study [41].
3.1.3. Transmission Techniques
Three commonly used transmission techniques for connected biomedical sensors are ZigBee,
Bluetooth Low Energy (BLE) and Wi-Fi. ZigBee is a two-way wireless communication technique
developed for sensors and control networks, which need a wider range, low latency, low energy
consumption at lower data rates. BLE is an alternative to the classical Bluetooth with higher data rate
and low power consumption within a limited area with low latency at 2.4 GHz. Wi-Fi makes a good
candidate for transmitting data with a data rate of up to 450 Mbps for indoor applications. However, it
imposes latency on the system of more than 25 ms and higher power consumption. Table 1 shows a
comparison between three different communication interfaces.
Table 1. Transmission methods comparison [42–46].
Latency
Speed
Power Consumption
Range
ZigBee
Bluetooth Low
Energy
15 ms
250 Kbps
9.3 mA
291 m
6 ms
1–11 Mbps
4.5 mA
10 m
Wi-Fi
≥25 ms
1.3 Gbps over 5
GHz and 450 Mbps
over 2.4 GHz
35 mA
50 m
Since the smart insole was intended for indoor application, BLE and WiFi both were suitable for
communication interface; however, the higher power consumption and latency made WiFi non-suitable
for smart insole application. Moreover, Simblee MCU has in-built BLE in its small form factor.
Therefore, BLE has been chosen as communication interface.
Sensors 2020, 20, 957
7 of 25
3.1.4. Power Management Unit (PMU)
Power supplies were chosen depending on the operating voltage of the system components. The
microcontroller and multiplexer both can operate at 3.3 V. The power management unit (PMU) is
LiPo Charger/Booster module MCP73831 [47] and AMS1117 voltage regulator connected to a Lithium
Polymer (LiPo) battery of 3.7 V (1000 mAh), which was regulated to 3.3 V. The PMU is not only
delivering regulated 3.3 V to the system but also capable to charge LiPo battery.
3.1.5. Host Computer
The acquired data from smart insole can be sent wirelessly to a host computer, where post
processing, thereby displaying the vGRF as pressure maps during gait cycle, was carried out.
The obtained data can be used in different gait analysis applications such as medical diagnostics,
rehabilitation and athlete’s performance assessment.
3.2. Sensors’ Calibration
The first step in designing the smart insole is to calibrate the force sensors that are going to be
used to detect the vGRF during the gait cycle. Three different force sensors were calibrated: FSR [20],
piezo-electric sensor [21] and piezo -vibration sensor [22].
3.2.1. Force-Sensitive Resistor (FSR) Calibration
Firstly, a voltage divider circuit must be used with the sensor to convert the resistance change (due
to applied force) of the sensor to a voltage value, which can be acquired by microcontrollers. Secondly, a
load cell of 5kg from HT sensor technology company [48], with HX711 amplifier modules [15] was used
as a weight reference for FSR calibration (Figure 3). The load cells consist of straight metal bar with two
strain gauge sensors and two normal resistors arranged in a Whitestone bridge configuration, a constant
excitation voltage (3–5 V) can be applied as an input to the circuit and the balanced configuration of
the circuit replicates a zero output voltage in normal conditions when no force is applied. Any force
applied to the load cell results in an unbalanced condition of the bridge leading to small voltage values
in the output that can be detected and converted to force [48]. The load cell has high sensitivity and
can detect as small as 1 gm of weight variation. However, the output voltage from the load cell is very
small, with a maximum value of 5 mV. Therefore, a HX711 amplifier module was used. The amplifier
module has instrumentational amplifier to amplify the signal with a 24-bit ADC that converts the
analog signal from the load cell bridge to digital value that is readable by a microcontroller. The HX711
transmits data to the microcontroller using I2 C communication protocol with 10Hz sampling rate [15].
The bar-type load cell was mounted with screws and spacers so that the strain can be measured
correctly (refer Figure 3B). The load cell was placed between two plates with only one side screwed
into each plate/board. This setup provides a moment of force on the strain gauges rather than just a
single compression force, resulting in higher sensitivity to applied forces. The output voltage from the
load cell exhibits a linear relationship with the applied force. This can be calibrated easily with any
small object of known mass such as a coin that weighs a few grams.
A known weight object (ex. a coin) was placed on load cell plate; the calibration factor was
adjusted until the output reading matches the known weight. Once the correct calibration factor is
obtained, it was used to convert the load cell voltages to corresponding weights. The calibration factor
is the slope of output voltage of load cell vs. real weights’ graph. The FSR was attached to adhesive
material on the back face of the active area, which was used to fix the FSR on the scale. A cylindrical
acrylic of 12.7 mm diameter, matching the active area of FSR, was used to apply force on the sensor
only. In addition, a square shaped acrylic plate was glued on top of the cylindrical acrylic to support
the weights, as shown in Figure 3. Then, 500 g weights are placed every 4 to 5 s until 5000 g is reached.
Readings from load cell and FSR circuit are acquired simultaneously by Arduino, which were saved in
a text file in a computer.
configuration, a constant excitation voltage (3–5 V) can be applied as an input to the circuit and the
balanced configuration of the circuit replicates a zero output voltage in normal conditions when no
force is applied. Any force applied to the load cell results in an unbalanced condition of the bridge
leading to small voltage values in the output that can be detected and converted to force [48]. The
load cell
Sensors
2020,has
20, high
957 sensitivity and can detect as small as 1 gm of weight variation. However, the output
8 of 25
voltage from the load cell is very small, with a maximum value of 5 mV. Therefore, a HX711 amplifier
module was used. The amplifier module has instrumentational amplifier to amplify the signal with
Finally,
the
FSR
output voltage
wassignal
plotted
withthe
respect
to the
load-cell
weight
and athat
mathematical
a 24‐bit
ADC
that
converts
the analog
from
load cell
bridge
to digital
value
is readable
relationship
was
derived.
The
equation
was
used
to
convert
smarts
insole
FSR
readings
into the
2
by a microcontroller. The HX711 transmits data to the microcontroller using I C communication
corresponding
applied
pressure
by[15].
the foot.
protocol with 10Hz
sampling
rate
A
Sensors 2020, 20, 957
8 of 25
B
Figure
3. FSR
calibration
setup
and
load‐cell
scale
Figure
3. FSR
calibration
setup
(A)(a)
and
load-cell
scale
(B).(b).
3.2.2. Piezo-Electric
Calibration
The bar‐typeSensor
load cell
was mounted with screws and spacers so that the strain can be measured
correctly
(refer
Figuremodule
3b). The
load
cellfor
was
placed
betweenwith
twosome
plates
with only one
screwed
The same
load-cell
was
used
piezo
calibration
modifications
(asside
shown
in
into
each
plate/board.
This
setup
provides
a
moment
of
force
on
the
strain
gauges
rather
than
just
Figure 4). Piezo transducers convert the applied mechanical forces into electrical impulses. Therefore, a
single
compression
force,(above
resulting
in higher
sensitivity
to applied
forces.
output
voltage
from the
a high
sampling
frequency
50 Hz)
is needed
to acquire
both the
piezoThe
output
and
the applied
load
cell
exhibits
a
linear
relationship
with
the
applied
force.
This
can
be
calibrated
easily
with
weights from the load cells. HX711 amplifier module samples the data with a low sampling frequencyany
small
of known
mass
such
as a coin
that weighs
a few ADC
grams.
of 10
Hz.object
Therefore,
the data
was
acquired
directly
by the 10-bit
of Arduino MCU with a sampling
A known
weight
object AD620AN
(ex. a coin)instrumentational
was placed on load
cell plate;
calibration
factorthe
was
frequency
of 1 kHz.
However,
amplifiers
[39]the
were
used before
adjusted
until
the
output
reading
matches
the
known
weight.
Once
the
correct
calibration
factor
is
acquisition step to amplify the small load cell outputs (maximum of 5 mV).
obtained,
it
was
used
to
convert
the
load
cell
voltages
to
corresponding
weights.
The
calibration
factor
Firstly, a voltage divider circuit was used to reduce the high piezo voltage outputs, which can go
slope
of output
loadcalibrated
cell vs. real
weights’
The FSR was
attached
adhesive
upistothe
20 V.
Secondly,
the voltage
load cellofwas
again
due tograph.
the modification.
Three
deadto
weights
material
on
the
back
face
of
the
active
area,
which
was
used
to
fix
the
FSR
on
the
scale.
A
cylindrical
of known masses were used: 500 kg, 2500 kg 5000 kg (maximum load for load cell). AD620AN
acrylic of 12.7 mm
diameter,
matching
the active
area
FSR, was
used towhen
applymaximum
force on the
sensor
instrumentational
amplifier
gain
was adjusted
to give
anofoutput
of voltage
load
is
only.
In
addition,
a
square
shaped
acrylic
plate
was
glued
on
top
of
the
cylindrical
acrylic
to
support
applied. This ensures that the full range of the Arduino ADC was utilized. Three dead weights were
the weights,
as shown
Figure
Then,
500 gvoltage
weights
are placed
every
4 toacquired
5 s until by
5000
g is reached.
added
one by one
on thein
scale
and3.the
output
from
load cells
were
Arduino.
A
Readings
from load
andbetween
FSR circuit
acquired
simultaneously
by Arduino,
were saved in
linear
relationship
wascell
fitted
theare
load
cell voltages
and applied
weights. which
This relationship
a text
a computer.
was
usedfile
to in
convert
the load cell voltage to a corresponding weight.
Finally,
the FSR
output
voltage
wastoplotted
with
to thesince
load‐cell
Unlike the FSR,
weights
cannot
be used
calibrate
the respect
piezo sensor,
piezo weight
sensors and
are a
mathematical
relationship
wasTherefore,
derived. The
was and
usedrelease
to convert
smarts insole
FSR readings
sensitive
to dynamic
forces only.
a fastequation
finger press
is suggested
as an alternative.
into
the
corresponding
applied
pressure
by
the
foot.
The calibration can be done by pressing the active area/ceramic of the sensor with various strengths
3.2.2. Piezo‐Electric Sensor Calibration
The same load‐cell module was used for piezo calibration with some modifications (as shown in
Figure 4). Piezo transducers convert the applied mechanical forces into electrical impulses. Therefore,
a high sampling frequency (above 50 Hz) is needed to acquire both the piezo output and the applied
Sensors 2020, 20, 957
9 of 25
and recording the generated electrical signals for each press as shown in Figure 4. Readings from the
load cell and piezo voltage divider circuit were acquired simultaneously by the Arduino MCU. Serial
terminal software was used to store the data in the computer. Load cell readings were plotted against
the piezo output voltage and a linear relationship was derived. The equation was used to convert
smarts
insole
readings into the corresponding applied weight by the foot.
Sensors 2020,
20, 957
9 of 25
Figure
Figure 4.
4. Piezoelectric
Piezoelectric sensor
sensor calibration
calibration setup.
setup.
3.2.3. Micro-Electromechanical Systems (MEMS) Sensor Calibration
Firstly, a voltage divider circuit was used to reduce the high piezo voltage outputs, which can
MEMS
producethe
alternating
(AC) impulses
and negative
go up
to 20sensors
V. Secondly,
load cell current
was calibrated
again with
due both
to thepositive
modification.
Threepeaks.
dead
Therefore,
modification
was required
setup
piezo
vibration
refer
to
weights oflittle
known
masses were
used: 500for
kg,the2500
kgof5000
kgelectric
(maximum
loadsensor,
for load
cell).
Figure
5.
An
offset
circuit
was
added
for
the
piezo-electric
acquisition
circuit.
The
piezo-vibration
AD620AN instrumentational amplifier gain was adjusted to give an output of voltage when
output
voltage
reducedThis
by aensures
voltagethat
divider
circuit
to ±1/2
wasThree
used
maximum
loadwas
is applied.
the full
range
of theVArduino
ADC amplifier
was utilized.
cc , then adder
to
add
an
offset
of
+1/2
V
,
so
the
new
AC
signal
will
be
centered
around
+1/2
V
with
maximum
dead weights were added
cc one by one on the scale and the output voltage from
cc load cells were
value
of Vby
minimum
of 0 V.
After modifying
the acquisition
circuit,
electric
acquired
Arduino.
A linear
relationship
was fitted
between the
load the
cell piezo
voltages
and sensors’
applied
cc and
calibration
steps
were
used
to
calibrate
the
piezo-vibration
sensor.
weights.
This
relationship
was
used
to
convert
the
load
cell
voltage
to
a
corresponding
weight.
Sensors 2020, 20, 957
10 of 25
Unlike the FSR, weights cannot be used to calibrate the piezo sensor, since piezo sensors are
sensitive to dynamic forces only. Therefore, a fast finger press and release is suggested as an
alternative. The calibration can be done by pressing the active area/ceramic of the sensor with various
strengths and recording the generated electrical signals for each press as shown in Figure 4. Readings
from the load cell and piezo voltage divider circuit were acquired simultaneously by the Arduino
MCU. Serial terminal software was used to store the data in the computer. Load cell readings were
plotted against the piezo output voltage and a linear relationship was derived. The equation was
used to convert smarts insole readings into the corresponding applied weight by the foot.
3.2.3. Micro‐Electromechanical Systems (MEMS) Sensor Calibration
MEMS sensors produce alternating current (AC) impulses with both positive and negative
peaks. Therefore, little modification was required for the setup of piezo electric vibration sensor, refer
to Figure 5. An offset circuit was added for the piezo‐electric acquisition circuit. The piezo‐vibration
output voltage was reduced by a voltage divider circuit to ±1/2 Vcc, then adder amplifier was used to
add an offset of +1/2 Vcc, so the new AC signal will be centered around +1/2 Vcc with maximum value
of Vcc and minimum of 0 V. After modifying the acquisition circuit, the piezo electric sensors’
Figure 5.
MEMS
setup.
5. LDT0-028k
LDT0‐028k
MEMS sensor
sensor calibration
calibration
calibration steps were usedFigure
to calibrate
the piezo‐vibration
sensor. setup.
3.3. Insole Fabrication
Once the sensors were calibrated, these sensors were separately used to construct the smart
insole for vGRF detection during gait cycles. The FSR sensors and piezo‐electric sensors were chosen
to construct two different insoles. While the piezo‐vibration sensor was found not suitable for vGRF
detection, the reason of not selecting the piezo‐vibration sensor is discussed in a later section. As
shown in Figure 6, the most common place of the foot plane, where most of the pressure is exerted
during gait are the heel, metatarsal heads, hallux and toe.
Sixteen sensors were placed on each insole to record pressure values in these areas. While no
sensors were placed on the medial arch area of the foot as most people exert very low/no pressure on
Sensors
2020,due
20, 957
10 of 25
that area
to it is arch shape [49]. Smart insole data were collected from 16 FSRs/piezo‐electric
sensors. Sixteen inputs were multiplexed to one output through a 16‐to‐1 multiplexer and applied to
an ADC input of the microcontroller then sent to host computer. All subjects were asked to place the
3.3. Insole Fabrication
sensor’s insole inside their shoes, then placing another layer of insole on top of it to ensure comfort
of the
subject
while walking.
The acquisition
and were
transmission
were
connected
through
a
Once
the sensors
were calibrated,
these sensors
separatelycircuit
used to
construct
the smart
insole
conductive
pathwayduring
that can
in minimizing
the size
of piezo-electric
the wire and sensors
avoidingwere
anychosen
electrical
for vGRF detection
gaithelp
cycles.
The FSR sensors
and
to
hazard.
were
worn While
by thethe
subject
inside his/her
ownwas
shoe
while
acquisition
and
constructThe
twoinsoles
different
insoles.
piezo-vibration
sensor
found
notthe
suitable
for vGRF
transmission
placed
inside
× 7cm box attached
the subject’s
by an
adhesive
detection, thecircuits
reasonwere
of not
selecting
thea 7cm
piezo-vibration
sensor istodiscussed
in aleg
later
section.
As
strap
belt
thecommon
data. Acquired
data
sent viawhere
Bluetooth
computer,
they
shown
in while
Figureacquiring
6, the most
place of
thewere
foot plane,
mostto
ofathe
pressurewhere
is exerted
were
plotted
and
during
gait are
theanalyzed.
heel, metatarsal heads, hallux and toe.
Figure 6. Area
Area of
of foot
foot selected
selected for sensors (A), and array of pressure sensor (B) in those areas.
sensors
were placed on each insole to record pressure values in these areas. While no
3.3.1.Sixteen
FSR Insole
Characterization
sensors were placed on the medial arch area of the foot as most people exert very low/no pressure
on that area due to it is arch shape [49]. Smart insole data were collected from 16 FSRs/piezo-electric
sensors. Sixteen inputs were multiplexed to one output through a 16-to-1 multiplexer and applied to
an ADC input of the microcontroller then sent to host computer. All subjects were asked to place the
sensor’s insole inside their shoes, then placing another layer of insole on top of it to ensure comfort
of the subject while walking. The acquisition and transmission circuit were connected through a
conductive pathway that can help in minimizing the size of the wire and avoiding any electrical hazard.
The insoles were worn by the subject inside his/her own shoe while the acquisition and transmission
circuits were placed inside a 7cm × 7cm box attached to the subject’s leg by an adhesive strap belt while
acquiring the data. Acquired data were sent via Bluetooth to a computer, where they were plotted
and analyzed.
3.3.1. FSR Insole Characterization
Twelve healthy subjects (Table 2) were asked to walk a straight 10 m walkway with self-selected
cadence six times with an average walking speed of 3–4 mile per hour (MPH) and data acquired
at 60 Hz sampling frequency using the smart insole made up of 16 FSRs (Figure 7). On treadmills,
participants are restricted to walk in straight line as direction changes and turning cannot be realized;
however, in the proposed study, the user walked freely in a 10-m walkway and they were asked to walk
in a corridor which has a length of 10m and width of 1.5 m and they did not need to walk completely in
a straight path and the user can walk in self-cadence, which is not possible on a treadmill. Subjects were
asked to place the smart insole in their shoe while wearing cotton socks to avoid any sweat leakage
that might damage the sensors or affect data acquisition from the sensors. Although walking speed is
an important factor in some applications, it is not needed in many gait studies where the main focus is
to detect the vertical ground reaction forces and asses the gait variables. The statistical gait variables
were the symmetry between both feet, percentage of different phases (stance and swing phase) and
sub phases (heel strike, mid-stance, toe off etc.) in a full gait cycle. Those statistical variables were
used in various studies including sports or medical applications for gait analysis, without the need for
participants are restricted to walk in straight line as direction changes and turning cannot be realized;
walk in a corridor which has a length of 10m and width of 1.5 m and they did not need to walk
however, in the proposed study, the user walked freely in a 10‐m walkway and they were asked to
completely in a straight path and the user can walk in self‐cadence, which is not possible on a
walk in a corridor which has a length of 10m and width of 1.5 m and they did not need to walk
treadmill. Subjects were asked to place the smart insole in their shoe while wearing cotton socks to
completely in a straight path and the user can walk in self‐cadence, which is not possible on a
avoid any sweat leakage that might damage the sensors or affect data acquisition from the sensors.
treadmill. Subjects were asked to place the smart insole in their shoe while wearing cotton socks to
Although walking speed is an important factor in some applications, it is not needed in many gait
avoid
any sweat leakage that might damage the sensors or affect data acquisition from the sensors. 11 of 25
Sensors 2020,
20, 957
studies
where the main focus is to detect the vertical ground reaction forces and asses the gait
Although walking speed is an important factor in some applications, it is not needed in many gait
variables. The statistical gait variables were the symmetry between both feet, percentage of different
studies where the main focus is to detect the vertical ground reaction forces and asses the gait
phases (stance and swing phase) and sub phases (heel strike, mid‐stance, toe off etc.) in a full gait
variables. The statistical gait variables were the symmetry between both feet, percentage of different
walkingcycle.
speed
measurement.
However,
the walking
recorded
the impact
of walking
Those
statistical variables
were used
in various speed
studieswas
including
sportsto
orsee
medical
applications
phases (stance and swing phase) and sub phases (heel strike, mid‐stance, toe off etc.) in a full gait
speed inforthe
forwithout
a gait cycle.
The
data
were
converted into
forcethe
values
byspeed
the relationship
gaitvGRF
analysis,
the need
for FSR
walking
speed
measurement.
However,
walking
was
cycle. Those statistical variables were used in various studies including sports or medical applications
recorded
to see the impact
of walking
speed
in thedata
vGRFwere
for a gait
cycle.
FSR
datainstance
were converted
obtained
theanalysis,
calibration
stage.
Then
sensors’
added
atThe
each
time
towas
obtain one
forin
gait
without
the need
for16
walking
speed measurement.
However,
the
walking speed
into
force
values
by
the
relationship
obtained
in
the
calibration
stage.
Then
16
sensors’
data were
value that
represents
theimpact
full force
exerted
byinthe
while
walking
(i.e.,
recorded
to see the
of walking
speed
thebody
vGRF for
a gait
cycle. The
FSRvGRF).
data were converted
added at each time instance to obtain one value that represents the full force exerted by the body
into force values by the relationship obtained in the calibration stage. Then 16 sensors’ data were
while walking (i.e., vGRF).
added at each time instance
to 2.
obtain
one value variables
that represents
the full force exerted by the body
Table
Demographic
of participants.
while walking (i.e., vGRF).
Table 2. Demographic variables of participants.
Body Mass
Number of
Gender
Age (Year)
Weight (kg)
Height (cm)
2
Index
(kg/m
Subjects
Table 2. Demographic variables of participants.
Body
Mass
Index)
Gender
Number of Subjects Age (Year) Weight (kg) Height (cm)
2)
7
30.1 ± 13.1
77.3 ± 21.2
159.8 ± 4.9 Body (kg/m
30.3
7.9
Female
Mass±Index
Gender
Number
of
Subjects
Age
(Year)
Weight
(kg)
Height
(cm)
7
30.1
±
13.1
77.3
±
21.2
159.8
±
4.9
30.3
±
7.9
FemaleMale
5
52.3 ± 4.8
83.5 ± 3.34
172.7 ± 11.7
28.3 ±2) 2.9
(kg/m
5
52.3 ± 4.8
83.5 ± 3.34
172.7 ± 11.7
28.3 ± 2.9
Male
7
30.1 ± 13.1
77.3 ± 21.2
159.8 ± 4.9
30.3 ± 7.9
Female
5
52.3 ± 4.8
83.5 ± 3.34
172.7 ± 11.7
28.3 ± 2.9
Male
Figure 7. Smart Insole using FSR sensor: top (A) and bottom (B).
Figure
7. Smart Insole using FSR sensor: top (A) and bottom (B).
Figure 7. Smart Insole using FSR sensor: top (A) and bottom (B).
3.3.2. Piezo‐Electric
Insole
Characterization
3.3.2. Piezo-Electric
Insole
Characterization
3.3.2.A
Piezo‐Electric
Insole
Characterization
similar test was
carried
out with the piezo‐electric sensor based smart insoles (Figure 8). Three
A similar
test was carried out with the piezo-electric sensor based smart insoles (Figure 8). Three
subjects were asked to walk in a 10 m walkway in normal cadence, with three trials carried out by
A similar
test
was carried
the piezo‐electric
sensor
based
smart
insoles (Figure 8). Three
subjectseach
were
asked
to walk
in a acquired
10out
mwith
walkway
in normal
cadence,
with
subject. The
data were
with a sampling
frequency
of 60
Hz. three trials carried out by each
subjects were asked to walk in a 10 m walkway in normal cadence, with three trials carried out by
subject. The data were acquired with a sampling frequency of 60 Hz.
each subject. The data were acquired with a sampling frequency of 60 Hz.
Figure 8. (A) Piezo insole with 16 piezo sensors, (B) additional insole layer placed on top on piezo
insole to ensure comfortability.
(A) Piezo
insole
with
piezosensors,
sensors, (B)
insole
layerlayer
placed
on topon
on top
piezo
Figure Figure
8. (A) 8.Piezo
insole
with
16 16
piezo
(B)additional
additional
insole
placed
on piezo
insole to ensure comfortability.
insole to ensure comfortability.
3.4. Performance Evaluation of the Prototype System
A commercial F-scan smart insole system (Figure 9A) was used to validate the designed insole.
The F-scan system is one of the best insoles currently available on the market. The insole comes with
ultra-thin (0.18 mm) flexible printed circuit with 960 sensing nodes. Each sensing element was recorded
with 8-bit resolution with a scanning speed up to 750 Hz. However, the overall cost of the system is
13,000 $ for the wired system and 17,000 $ for the wireless system. On the other hand, the instrumented
insole costs only ~500 $. Usually, the vGRF peak is around ±10% of the subject’s weight. Therefore,
using F-scan software, data collected from each subject was calibrated based on subject’s weight. The
user needs to stand on one foot applying his/her full weight on the insole for 4 to 5 s, then the average
applied weight was calculated. If the value obtained was less than the subject’s weight, the F-scan
software adjusted the output by a multiplication factor. Similar approach was used in the prototyped
FSR insole as well. Figure 9B,C show the F-scan and prototyped system worn by the same subject to
compare the vGRF signal acquired by the individual system.
instrumented insole costs only ~500 $. Usually, the vGRF peak is around ±10% of the subject’s weight.
Therefore, using F‐scan software, data collected from each subject was calibrated based on subject’s
weight. The user needs to stand on one foot applying his/her full weight on the insole for 4 to 5 s,
then the average applied weight was calculated. If the value obtained was less than the subject’s
weight, the F‐scan software adjusted the output by a multiplication factor. Similar approach was used
Sensors 2020,
20, 957
in the
prototyped FSR insole as well. Figure 9B,C show the F‐scan and prototyped system worn by 12 of 25
the same subject to compare the vGRF signal acquired by the individual system.
F‐scan commercial system (A), F‐scan system worn by Subject 01 (B) and FSR‐based
Figure 9.Figure
F-scan9.commercial
system (A), F-scan system worn by Subject 01 (B) and FSR-based prototype
prototype system worn by Subject 01 (C).
system worn by Subject 01 (C).
4. Analysis
4. Analysis
This section explains the mathematical calculations and analyses used for the sensor calibrations
Thisand
section
the mathematical calculations and analyses used for the sensor calibrations
insole explains
characterization.
and insole characterization.
4.1. Sensors’ Calibration
4.1. Sensors’ Calibration
4.1.1. FSR Sensor Calibration
4.1.1. FSR Sensor
Calibration
The FSR
sensors exhibits resistance change in correspondence to the applied force. Therefore, a
divider circuit was used to convert the resistance changes to voltage values to be acquired by
Thevoltage
FSR sensors
exhibits resistance change in correspondence to the applied force. Therefore, a
microcontroller.
voltage divider circuit was used to convert the resistance changes to voltage values to be acquired
𝑅
11 𝑘𝛺
by microcontroller.
(1)
𝑉
𝑉
5𝑉
𝑅 𝐹𝑆𝑅
11 𝑘𝛺 𝐹𝑆𝑅
R
11 kΩ
Vout
= VCCthe
× FSR resistance
= 5V
× decreases, showing an increased output
(1)
As the applied force
increases,
also
R + FSR
11 kΩ + FSR
voltage according to Equation (1). The acquired voltages were then converted to their equivalent FSR
As the
applied
force
increases, the
FSR resistance
also decreases, showing an increased output
resistance
values
by substitution
of Equation
(1).
voltage according to Equation (1). The acquired
voltages
were then converted to their equivalent FSR
5𝑉
11𝑘𝛺
11𝑘𝛺
𝐹𝑆𝑅 (1).
(2)
resistance values by substitution of Equation
𝑉
FSR =
4.1.2. Piezo‐Electric Sensor Calibration
5V × 11kΩ
− 11kΩ
Vout
(2)
The piezo‐electric sensor generates high voltage values, as high as 20 V with weights less than 5
kg, which requires
using
a voltage divider circuit before data acquisition by microcontroller.
4.1.2. Piezo-Electric
Sensor
Calibration
V
𝑉𝑜𝑙𝑡𝑎𝑔𝑒
𝑑𝑖𝑣𝑖𝑑𝑒𝑟
𝐺𝑎𝑖𝑛 as Vhigh as 20 V with weights less
(3) than 5
The piezo-electric sensor generates
high
voltage
values,
kg, which requires using a voltage divider circuit before data acquisition by microcontroller.
Vmax input = Voltage divider Gain × VPiezo max
⇒ Voltage divider Gain =
Vmax input
VPiezo max
=
VCC
VPiezo max
=
5V
= 0.25
20V
(3)
(4)
Therefore, the voltage divider circuit were chosen as follows:
Vout =
R1
3 MΩ
× VPiezo =
× VPiezo = 0.25VPiezo
R1 + R2
3 MΩ + 9 MΩ
(5)
Substituting the maximum piezo voltage in Equation (5) gives:
Vout max = 0.25VPiezo max = 0.25(20V ) = 5V
(6)
This ensures that maximum microcontroller input voltage (5 V) was not exceeded. The acquired
voltages were then converted to their equivalent Piezo sensor voltage outputs by subject substitution
of Equation (5).
1
VPiezo =
× Vout = 4Vout
(7)
0.25
Sensors 2020, 20, 957
13 of 25
There are different equivalent electrical models for the piezo-electric sensors [28]. A simplified
common model is a voltage source/generator with a capacitance, which was used in this study. Usually,
the capacitance values are in Nano Farad range. The equivalent capacitance is typically measured
using a parallel connection of a capacitance meter to the sensor. Connecting the piezo-electric sensor to
the voltage divider circuit forms a first order high-pass filter. Therefore, high resistance values in mega
ohms were used to ensure that most of the generated frequencies by the applied forces would pass.
Assuming equivalent capacitance of piezo-electric sensor equal to 9 nF, the cut-off frequency can be
written as:
1
1
fcutt−o f f =
=
= 1.47Hz
(8)
2πRC
2π(3M + 9M)9nF
Apart from DC and very low frequency components, other signal components were expected to
be applied to the MCU input. AD620AN instrumentational amplifiers were used to amplify the low
amplitude load cell signals, before it was applied to the microcontroller. The load cells give an output
of maximum 40 mV, which can be amplified to the full-scale range of the analog channel. Therefore,
the gain of the amplifier and the amplifier gain resistor were chosen as follows:
G=
Vcc
Vload max
RG =
=
5V
= 125
40mV
49.9kΩ
49.9k
=
= 402 Ω
G−1
124
(9)
(10)
4.1.3. MEMS Sensor Calibration
As mentioned previously, the MEMS generates positive or negative amplitude signals based on
the applied force in x, y or z directions. This requires an offset circuit along with a voltage divider
circuit to reduce the signal amplitude. It is assumed the piezo-vibration output can go up to 10 V with
the maximum applied force.
Vmax input = Voltage divider Gain × VPiezo max
Voltage divider Gain =
Vmax input
VPiezo max
=
VCC /2
5V/2
=
= 0.25
VPiezo max
10V
(11)
(12)
Therefore, the voltage divider circuit were chosen as follows:
Vout =
R1
3 MΩ
× VPiezo =
× VPiezo = 0.25VPiezo ,
R1 + R2
3 MΩ + 9 MΩ
(13)
Substituting the maximum and minimum piezo voltage in Equation (13) gives:
Vout max/min = 0.25VPiezo max/min = 0.25(±10V ) = ±2.5V
(14)
The next step is to add an offset of 1/2Vcc to ensure that the signal was within 0 V to Vcc range.
4.2. Piezo-Electric Sensor Response
The piezo-electric sensors can detect the applied forces efficiently, by converting the mechanical
movements into electrical signals. However, the movements need to be dynamic. The piezo-electric
sensor generated electrical pulses that mimicked the applied mechanical movement. If the mechanical
movement was a fast press and release of finger on the active area of the piezo-electric sensor, the pulse
was shrunk to an impulse-liked shape.
On the other hand, if a gentle force was applied by a slow press and remove by the palm of a
hand, the generated signal had irregular pulse shape with longer duration compared to the fast finger
press. Even though the piezo-electric sensor’s output can mimic dynamically changing force, it fails to
Sensors 2020, 20, 957
14 of 25
The
Sensors 2020, 20,
957next step is to add an offset of 1/2Vcc to ensure that the signal was within 0 V to Vcc range.
14 of 25
4.2. Piezo‐Electric Sensor Response
detect a static
force.
Therefore,
when
applied
forceforces
is a mixture
and
force such as
The
piezo‐electric
sensors
can the
detect
the applied
efficiently,ofbydynamic
converting
the static
mechanical
movements
into electrical
signals. However,
the movements
beacquire
dynamic.static
The piezo‐electric
smart insole
application,
the piezo-electric
sensors
cannot beneed
usedtoto
pressure. However,
sensor generated
electrical
that
mimicked
the applied
movement.
If theFigure 10
the piezo-electric
sensors
can bepulses
used to
detect
heel strike
or toemechanical
off with good
accuracy.
mechanical movement was a fast press and release of finger on the active area of the piezo‐electric
illustrates the individual sensor output for the different applied forces.
sensor, the pulse was shrunk to an impulse‐liked shape.
Figure 10.
Piezo
insoleinsole
sensors
output
serial
plotter:
(A)finger
fast finger
press
and release,
Figure
10. Piezo
sensors
outputwith
with Arduino
Arduino serial
plotter:
(A) fast
press and
release,
slowpress
palm press
and release,
sensorsoutput
output for
walking.
(B) slow(B)
palm
and release,
(C)(C)
sensors
fortwo‐step
two-step
walking.
Onshows
the other
if a from
gentleaforce
was
applied by a slow
press
andinsole
remove
the gait
palmcycles.
of a
Figure 11
thehand,
output
single
piezo-electric
sensor
of an
forbyfew
When
hand, the generated signal had irregular pulse shape with longer duration compared to the fast finger
a force was applied vertically on the sensor’s active area (ceramic), it compressed, exerting an electrical
press. Even though the piezo‐electric sensor’s output can mimic dynamically changing force, it fails
impulse to
with
a positive
peakTherefore,
that mimicked
mechanical
applied.
Theand
electrical
signal
went back
detect
a static force.
when thethe
applied
force is aforce
mixture
of dynamic
static force
such
to zero. as
Assmart
the applied
force wasthe
released,
the signal
continued
some
negative
values
as the piezo
insole application,
piezo‐electric
sensors
cannot be to
used
to acquire
static
pressure.
piezo‐electric
can of
be the
usedapplied
to detectforce.
heel strike
or toe
with good
accuracy.
ceramic However,
bouncedthe
to the
oppositesensors
direction
Finally,
theoffsignal
returned
to zero. The
Figure
10
illustrates
the
individual
sensor
output
for
the
different
applied
forces.
microcontroller clipped the negative part of the signal. However, some part of the negative signal was
Figure 11 shows the output from a single piezo‐electric sensor of an insole for few gait cycles.
still there, due to the offset added in the acquisition circuit as illustrated in Figure 11.
a force
SensorsWhen
2020, 20,
957 was applied vertically on the sensor’s active area (ceramic), it compressed, exerting an15 of 25
Voltage
Voltage
Voltage
electrical impulse with a positive peak that mimicked the mechanical force applied. The electrical
signal
went back to zero. As the applied force was released, the signal continued to some negative
15V
values as the piezo ceramic bounced to the opposite direction of the applied force. Finally, the signal
returned to zero. The microcontroller clipped the negative part of the signal.
1.3V However, some part of
1V
the negative signal was still there, due to the offset added in the acquisition circuit as illustrated in
Figure 11.
0.3V
0
t
‐4.8V
0
‐0.32V
Example of a piezo
electric sensors output
when a vertical force is
applied then released
‐0.02V
t
t
Added offset (0.2‐0.4V), this could be
added noise picked by the high impedance
resistors of the voltage divider (in Mega
ohms range)
The piezo signal after being
attenuated by the voltage divider circuit,
assuming a factor of (1/15)
0.3V
t
0
The analog input to the
microcontroller. The negative
part of the signal will be clipped
Voltage
1V
1.3V
Voltage
Voltage
15V
0
t
‐0.30V
Removing
the offset (‐0.3)
0
t
‐4.5V
Multiplying by a gain
(x15) to compensate for the
attenuation by the voltage
divider circuit
Voltage
The signal will repeat for the next gait cycles, generating a
periodic signal. Normally the output during the swing phase
is expected to be zero. However, the swing phase intersect
with the negative part following the stance phase. That’s why
no zero period is shown between the consecutive cycles
t
Figure11.
11.Mimicking
Mimickingpiezoelectric
piezoelectric sensor
sensor output
Figure
outputduring
duringgait
gaitcycle.
cycle.
5. Results and Discussion
This section illustrates and discuss the results obtained from calibration and characterization
tests.
Sensors 2020, 20, 957
15 of 25
5. Results and Discussion
This section illustrates and discuss the results obtained from calibration and characterization tests.
5.1. Sensors’ Calibration
5.1.1. FSR Sensor Calibration
Weight [kg]
Weight [kg]
Output Voltage [V]
Three calibration trials were undertaken for one FSR sensor from Interlink Electronics [22],
following the calibration procedure explained previously; 500 g weights where placed one by one every
3–4 s until it reached 5000 g, followed by unloading process from 5000 g down to 0 g. In the loading
experiment, output voltage from the voltage divider circuit showed increasing values reflecting the
decrease in FSR resistance as shown in Figure 12A. When the applied weight was constant, the output
Sensors
2020,
20, 957 constant with small variations.
16 of 25
voltage
remained
(A)
(B)
Figure
Figure 12. FSR
FSR calibration
calibration test
test (A)
(A) applied
applied weight
weight and
and FSR
FSR circuit output vs. time
time (B)
(B) applied
applied weight
weight
FSR resistance.
resistance.
vs. FSR
Weight [kg]
In addition,
if the
constant
for a longer
time (a few
minutes),
the sensor voltage
For
each of the
three
trials, weight
loadingwas
andkept
off‐loading
relationships
where
obtained.
stabilized
to atests
steady
value.
However, the aim of this study was to investigate the dynamic response of
Loading
best
fit relations:
.
the FSR. Therefore, the average
output voltage4233.3
for the∗ sample
were calculated
and plotted against the
𝑊𝑒𝑖𝑔ℎ𝑡
𝑅𝑒𝑠𝑖𝑠𝑡𝑎𝑛𝑐𝑒
.
corresponding applied weights.
Figure 12B 3657.4
shows the
plotted data with the fitted waveform. The
𝑊𝑒𝑖𝑔ℎ𝑡
∗ 𝑅𝑒𝑠𝑖𝑠𝑡𝑎𝑛𝑐𝑒
.
calibration showed slight difference
the loading
and unloading
curves, which was expected
𝑊𝑒𝑖𝑔ℎ𝑡between 5117.4
∗ 𝑅𝑒𝑠𝑖𝑠𝑡𝑎𝑛𝑐𝑒
due to the hysteresis behavior of FSRs. However, the error was caused by the FSR hysteresis, which
can be neglected, as the difference was not significantly high. This can be justified if the response from
the smart insole using FSR sensors resembles typical vGRF reported in the literature.
Off-Loading tests best fit relations:
WeightTrial1 = 5035.2 ∗ Resistance−1.72
WeightTrial2 = 3436.5 ∗ Resistance−1.895 WeightTrial3 = 8111.8 ∗ Resistance−2.589
It is evident that the first and third trial relationships were close to each other (Figure 13). Therefore,
either of them can be chosen for the FSR insole. The second trial showed a steaper curve due to higher
hystersis error.
Figure 13. FSR calibration test: best fit curves between applied weight and FSR resistance for three
trials.
Off‐Loading tests best fit relations:
𝑊𝑒𝑖𝑔ℎ𝑡
5035.2 ∗ 𝑅𝑒𝑠𝑖𝑠𝑡𝑎𝑛𝑐𝑒 .
𝑊𝑒𝑖𝑔ℎ𝑡
3436.5 ∗ 𝑅𝑒𝑠𝑖𝑠𝑡𝑎𝑛𝑐𝑒 . 𝑊𝑒𝑖𝑔ℎ𝑡
8111.8 ∗ 𝑅𝑒𝑠𝑖𝑠𝑡𝑎𝑛𝑐𝑒 .
It is evident that the first and third trial relationships were close to each other (Figure 13).
(A)
(B)
Sensors 2020,Figure
20, 957 12. FSR calibration test (A) applied weight and FSR circuit output vs. time (B) applied weight
16 of 25
vs. FSR resistance.
Weight [kg]
For each of the three trials, loading and off‐loading relationships where obtained.
Loading tests best fit relations:
𝑊𝑒𝑖𝑔ℎ𝑡
4233.3 ∗ 𝑅𝑒𝑠𝑖𝑠𝑡𝑎𝑛𝑐𝑒 .
𝑊𝑒𝑖𝑔ℎ𝑡
3657.4 ∗ 𝑅𝑒𝑠𝑖𝑠𝑡𝑎𝑛𝑐𝑒 .
𝑊𝑒𝑖𝑔ℎ𝑡
5117.4 ∗ 𝑅𝑒𝑠𝑖𝑠𝑡𝑎𝑛𝑐𝑒 .
Figure
13.calibration
FSR calibration
test:
fit between
curves between
appliedand
weight
and FSR for
resistance
for three
Figure
13. FSR
test: best
fit best
curves
applied weight
FSR resistance
three trials.
trials.
5.1.2. Piezo-Electric Sensor Calibration
bestwere
fit relations:
Two Off‐Loading
piezoelectrictests
sensors
used in the calibration process. Three trials were conducted on
.
𝑊𝑒𝑖𝑔ℎ𝑡
5035.2 The
∗ 𝑅𝑒𝑠𝑖𝑠𝑡𝑎𝑛𝑐𝑒
the first sensors with four trials for
the second sensor.
piezoelectric sensors showed .a linear
.
Sensors 2020, 20, 957
𝑊𝑒𝑖𝑔ℎ𝑡
3436.5 ∗ (Figure
𝑅𝑒𝑠𝑖𝑠𝑡𝑎𝑛𝑐𝑒
𝑊𝑒𝑖𝑔ℎ𝑡
8111.8 ∗ 𝑅𝑒𝑠𝑖𝑠𝑡𝑎𝑛𝑐𝑒 17 of 25
relationship with
the applied weights
14).
It is evident that the first and third trial relationships were close to each other (Figure 13).
Therefore, either of them can be chosen for the FSR insole. The second trial showed a steaper curve
Reference Mass vs Time
due to higher hystersis error.
6
4
5.1.2. Piezo‐Electric Sensor Calibration
2
Weight [kg]
Two piezoelectric sensors were used in the calibration process. Three trials were conducted on
the
sensor. The piezoelectric sensors showed a linear
0 first
2 sensors
4
6 with
8 four
10 trials
12 for
14 the
16 second
18
Time [sec]
relationship with the
applied weights (Figure 14).
0
Piezo Voltage vs Time
15
10
5
0
0
2
4
6
8
10
Time [sec]
12
14
16
18
(A)
(B)
Figure 14.
14. Piezo‐electric
calibration test:
test: (A)
weight vs.
vs. time
output
Figure
Piezo-electric sensor
sensor calibration
(A) applied
applied weight
time and
and piezoelectric
piezoelectric output
voltage
vs.
time
(B)
applied
weight
vs.
piezoelectric
output
voltage.
voltage vs. time (B) applied weight vs. piezoelectric output voltage.
trials for
for the
the 2nd
2nd piezo
The second and fourth trials
piezo sensor
sensor had
had different
different slopes compared to the
trials. This
This could
couldbe
berelated
relatedtotothe
thecalibration
calibrationprocess
process
itself,
weights
were
applied
remaining trials.
itself,
asas
thethe
weights
were
applied
by
by fast
presses
and
releases
active
area
thesensor.
sensor.Therefore,
Therefore,applying
applyingthe
theforce
force on
on the
the exact
fast
presses
and
releases
onon
thethe
active
area
ofofthe
guaranteed between successive readings. This
same areas is not guaranteed
This issue
issue can
can be overcome by using a
machine to apply the weights. However, this would defeat the purpose of the study in providing a
low‐cost setup (Figure 15).
low-cost
6
5
4
Weight vs Piezo-electric Voltage
remaining trials. This could be related to the calibration process itself, as the weights were applied
by fast presses and releases on the active area of the sensor. Therefore, applying the force on the exact
same areas is not guaranteed between successive readings. This issue can be overcome by using a
machine to apply the weights. However, this would defeat the purpose of the study in providing a
Sensors 2020, 20, 957
17 of 25
low‐cost setup (Figure 15).
Weight vs Piezo-electric Voltage
6
5
4
3
Piezo1 Trial1
Piezo1 Trial2
Piezo1 Trial3
Piezo2 Trial1
Piezo2 Trial2
Piezo2 Trial3
Piezo2 Trial4
2
1
0
-1
0
2
4
6
8
Voltage [V]
10
12
14
Figure
15. 15.
Piezoelectric
test:seven
seventrials
trials
shows
relationship
between
the applied
Figure
Piezoelectricsensor
sensorcalibration
calibration test:
shows
relationship
between
the applied
weight
piezoelectricoutput
output voltage.
voltage.
weight
andand
piezoelectric
Obtained lines of best fit:
Obtained
lines of best fit:
𝑊𝑒𝑖𝑔ℎ𝑡
𝑃𝑖𝑒𝑧𝑜𝑉𝑜𝑙𝑡𝑎𝑔𝑒
0.19123
WeightPiezo1Trial1 = 0.42867
0.42867 ∗ ∗PiezoVoltage
− 0.19123
𝑊𝑒𝑖𝑔ℎ𝑡
0.41110 ∗ 𝑃𝑖𝑒𝑧𝑜𝑉𝑜𝑙𝑡𝑎𝑔𝑒 0.0081012
𝑊𝑒𝑖𝑔ℎ𝑡
0.39321∗ PiezoVoltage
∗ 𝑃𝑖𝑒𝑧𝑜𝑉𝑜𝑙𝑡𝑎𝑔𝑒
0.084656
WeightPiezo1Trial2 = 0.41110
+ 0.0081012
𝑊𝑒𝑖𝑔ℎ𝑡
0.34619 ∗ 𝑃𝑖𝑒𝑧𝑜𝑉𝑜𝑙𝑡𝑎𝑔𝑒 0.4105
WeightPiezo1Trial3 = 0.39321 ∗ PiezoVoltage + 0.084656
𝑊𝑒𝑖𝑔ℎ𝑡
0.27242 ∗ 𝑃𝑖𝑒𝑧𝑜𝑉𝑜𝑙𝑡𝑎𝑔𝑒 0.57351
WeightPiezo2Trial1 =0.35564
0.34619 ∗∗PiezoVoltage
+ 0.4105
𝑊𝑒𝑖𝑔ℎ𝑡
𝑃𝑖𝑒𝑧𝑜𝑉𝑜𝑙𝑡𝑎𝑔𝑒
0.30325
𝑊𝑒𝑖𝑔ℎ𝑡
𝑃𝑖𝑒𝑧𝑜𝑉𝑜𝑙𝑡𝑎𝑔𝑒
0.36416
WeightPiezo2Trial2 = 0.31765
0.27242 ∗ ∗
PiezoVoltage
+ 0.57351
WeightPiezo2Trial3 = 0.35564 ∗ PiezoVoltage + 0.30325
5.1.3. MEMS Sensor Calibration
WeightPiezo2Trial4 = 0.31765 ∗ PiezoVoltage + 0.36416
5.1.3. MEMS Sensor Calibration
Twenty different calibration trials were conducted on a piezo-vibration sensor. However, high
repeatability error persisted, making it difficult to obtain a clear relation between sensor output voltage
and the applied weight. The applied weight showed a direct proportional relation with output voltage
for some successive readings and an inverse relation with some other successive readings. This is
because of the MEMS sensitivity to the applied force in 3-D space (x, y or z directions). It generates 1-D
output voltage with positive or negative amplitude depending on the applied force in certain direction.
Therefore, if the applied force is a summation of forces in 2 or 3 axes, the output voltage might go to
zero or attenuated with the addition of different sign amplitudes.
A linear relation was not clearly obtained by the application of vertical forces, as the applied force
might not be applied in one axis only (Figure 16).
readings. This is because of the MEMS sensitivity to the applied force in 3‐D space (x, y or z
directions). It generates 1‐D output voltage with positive or negative amplitude depending on the
applied force in certain direction. Therefore, if the applied force is a summation of forces in 2 or 3
axes, the output voltage might go to zero or attenuated with the addition of different sign amplitudes.
A2020,
linear
relation was not clearly obtained by the application of vertical forces, as the applied
Sensors
20, 957
18 of 25
force might not be applied in one axis only (Figure 16).
Sensors 2020, 20, 957 Reference Mass vs Time
18 of 25
4
Amplitude [V]
Weight [kg]
Twenty different calibration trials were conducted on a piezo‐vibration sensor. However, high
repeatability error persisted, making it difficult to obtain a clear relation between sensor output
2
voltage and the applied weight. The applied weight showed a direct proportional relation with
output
voltage for some successive readings and an inverse relation with some other successive
0
0
5
10
20
25
readings. This
is because
of15the MEMS
sensitivity
to the applied force in 3‐D space (x, y or z
Time [sec]
directions). It generates
1‐D
output
voltage
with
positive
or negative amplitude depending on the
Piezo-vibration Voltage vs Time
applied force in certain direction. Therefore, if the applied force is a summation of forces in 2 or 3
4
axes, the output voltage might go to zero or attenuated with the addition of different sign amplitudes.
A linear relation was not clearly obtained by the application of vertical forces, as the applied
2
force might not be applied in one axis only (Figure 16).
0
Weight [kg]
0
5
10
15
Reference
Mass vs Time
Time [sec]
4
20
25
(a)
(b)
2
Figure
16.
16. MEMS
MEMSsensor
sensor calibration
calibration test:
test: (A)
(A)applied
applied weight
weight vs.
vs. time
timeand
andMEMS
MEMSsensor
sensoroutput
output voltage
voltage
vs. time
time (B)
(B) applied
applied weight
weight vs.
vs. MEMS
MEMS sensor
sensor output
output voltage.
voltage.
0
0
5
0
5
10
15
20
25
20
25
0
Weight [kg]
Amplitude [V]
The mathematical relations
Time [sec] obtained in the calibration phase cannot be used to design a
piezo-vibration sensor-based
smart
insole, since the applied force in gait can be in any of the x,
Piezo-vibration Voltage
vs Time
y or z4directions (Figure 17). Therefore, the piezo-vibration sensor was discarded from the sensor list
for designing smart insole. However, it can be utilized in other biomedical applications where the
force 2directions are limited to a certain axis or a fixed plane. Moreover, it can be used to detect initial
timing of the applied force. The lines of best fit obtained were:
10
15
TimeWeight
[sec]
Trial1
= 1.4145 ∗ PiezoVoltage − 1.3447
(a)
WeightTrial2 = 1.5215 ∗ PiezoVoltage − 1.8581
(b)
Figure 16. MEMS sensor calibration test: (A) applied weight vs. time and MEMS sensor output voltage
WeightTrial3 = 0.80343 ∗ PiezoVoltage − 0.22721
vs. time (B) applied weight vs. MEMS sensor output voltage.
Figure 17. MEMS sensor calibration test: 3 best trials shows relationship between the applied weight
and MEMS sensor’s output voltage.
Weight [kg]
The mathematical relations obtained in the calibration phase cannot be used to design a piezo‐
vibration sensor‐based smart insole, since the applied force in gait can be in any of the x, y or z
directions (Figure 17). Therefore, the piezo‐vibration sensor was discarded from the sensor list for
designing smart insole. However, it can be utilized in other biomedical applications where the force
directions are limited to a certain axis or a fixed plane. Moreover, it can be used to detect initial timing
of the applied force. The lines of best fit obtained were:
Figure
MEMSsensor
sensorcalibration
calibrationtest:
test:3 3best
besttrials
trialsshows
showsrelationship
relationshipbetween
betweenthe
the
applied
weight
Figure
17.17.MEMS
applied
weight
and
MEMS
sensor’soutput
outputvoltage.
voltage.
and
MEMS
sensor’s
The mathematical relations obtained in the calibration phase cannot be used to design a piezo‐
vibration sensor‐based smart insole, since the applied force in gait can be in any of the x, y or z
directions (Figure 17). Therefore, the piezo‐vibration sensor was discarded from the sensor list for
designing smart insole. However, it can be utilized in other biomedical applications where the force
directions are limited to a certain axis or a fixed plane. Moreover, it can be used to detect initial timing
of the applied force. The lines of best fit obtained were:
Sensors 2020, 20, 957
Sensors 2020, 20, 957
19 of 25
𝑊𝑒𝑖𝑔ℎ𝑡
𝑊𝑒𝑖𝑔ℎ𝑡
𝑊𝑒𝑖𝑔ℎ𝑡
1.4145 ∗ 𝑃𝑖𝑒𝑧𝑜𝑉𝑜𝑙𝑡𝑎𝑔𝑒
1.5215 ∗ 𝑃𝑖𝑒𝑧𝑜𝑉𝑜𝑙𝑡𝑎𝑔𝑒
0.80343 ∗ 𝑃𝑖𝑒𝑧𝑜𝑉𝑜𝑙𝑡𝑎𝑔𝑒
1.3447
1.8581
0.22721
19 of 25
5.2.
Insole
Charecterization
5.2.
Insole
Charecterization
5.2.1.
FSR-Based
Insole
Characterization
5.2.1.
FSR‐Based
Insole
Characterization
The
gait
cycle
ofof
1212
subjects
were
recorded
while
walking
onon
a 10
mm
walkway
in in
self-selected
The
gait
cycle
subjects
were
recorded
while
walking
a 10
walkway
self‐selected
walking
manner.
Each
subject
had
6 trials
recorded,
where
the
first
and
last
few
(1(1
toto
3) 3)
cycles
were
walking
manner.
Each
subject
had
6 trials
recorded,
where
the
first
and
last
few
cycles
were
discarded
from
cycles for
forboth
bothfeet
feetwere
wereconsidered
consideredfor
discarded
fromeach
eachtrial,
trial,and
andthe
theremaining
remaining part
part of the gait cycles
foranalysis.
analysis.
The
cycle
of one
of the
subjects
is analyzed
in following
the following
illustrating
a
The
gaitgait
cycle
of one
of the
subjects
is analyzed
in the
lines,lines,
illustrating
a simple
simple
analysis
technique
replicated
different application
application by researchers
analysis
technique
that that
can can
be be
replicated
inindifferent
researchersworking
workingwith
with
wearable
insoles.
wearable
insoles.
InIn
normal
gait
cycles
both
heel
peak
(first
peak)
and
toe
offoff
(peak)
must
show
close
values,
with
normal
gait
cycles
both
heel
peak
(first
peak)
and
toe
(peak)
must
show
close
values,
with
both
feet
having
symmetrical
signals.
signal
showed
close
peak
values
both
feet
having
symmetrical
signals.Even
Eventhough
thoughthe
theright
rightfoot
foot
signal
showed
close
peak
values
(Figure
18),
thethe
left
foot
signals
showed
a big
variance
between
thethe
heel-strike
and
toe-off
peaks.
This
(Figure
18),
left
foot
signals
showed
a big
variance
between
heel‐strike
and
toe‐off
peaks.
This
can
bebe
explained,
the
sensitivity
between
the
FSRs
ofof
the
insole
and
their
hysteresis
can
explained,byby
the
sensitivitydifference
difference
between
the
FSRs
the
insole
and
their
hysteresis
effect.
AsAs
explained
earlier,
this issue
was mitigated
by some
teams using
a regression
models
effect.
explained
earlier,
this issue
was mitigated
byresearch
some research
teams
using a regression
that
calibrates
the FSR insole
readings
a reference
signal, recorded
in motionin
models
that calibrates
the FSR
insoleagainst
readings
against a reference
signal, simultaneously
recorded simultaneously
analysis
[34,35].
This
expensive
be neglected
some applications,
where the quality
motionlabs
analysis
labs
[35,50].
This approach
expensivecan
approach
can beinneglected
in some applications,
where
ofthe
thequality
acquired
is sufficient
toisachieve
theto
desired
goal.
instance,
detection
application,
ofsignal
the acquired
signal
sufficient
achieve
the For
desired
goal.smart
For instance,
smart
detection
where
the machine-learning
algorithm can differentiate
different groups
people even
withof
application,
where the machine‐learning
algorithm between
can differentiate
betweenofdifferent
groups
lowor medium-quality
gait cycles recorded
(vGRF). gait cycles (vGRF).
people
even with low‐ recorded
or medium‐quality
Gait Cycles
50
Left Foot
Right Foot
40
30
20
10
0
0
1
2
Time [sec]
3
4
Figure
Gait
cycles
readings
left
and
right
foot
with
FSR
smart
insole.
Figure
18.18.
Gait
cycles
readings
forfor
left
and
right
foot
with
FSR
smart
insole.
The
full
gait
record
was
segmented
into
distinct
gait
cycles.
Then
was
resampled
into
512
The
full
gait
record
was
segmented
into
distinct
gait
cycles.
Then
it it
was
resampled
into
toto
512
sample.Segmentation
Segmentationisisa acommon
commonpractice
practicetotofacilitate
facilitatethe
thecomparison
comparisonbetween
betweenallallthe
thegait
gaitcycles.
cycles.
sample.
The
segmented
gait
cycles
used
smart
detection
algorithms
where
segments
equal
length
are
The
segmented
gait
cycles
areare
used
inin
smart
detection
algorithms
where
segments
ofof
equal
length
are
used
train
specific
machine
learning
algorithms
to classify
different
groups
of people
based
used
to to
train
specific
machine
learning
algorithms
to classify
different
groups
of people
based on
theiron
their
In addition,
statistical
the segmented
such
as mean,
standard
deviation,
gait.
Ingait.
addition,
statistical
data ofdata
the of
segmented
cyclescycles
such as
mean,
standard
deviation,
timetime
to
to
peaks,
and
percentage
of
stance
phase
in
a
full
cycle/stride
(stance
phase
plus
swing
phase)
can
peaks, and percentage of stance phase in a full cycle/stride (stance phase plus swing phase) can bebe
utilized
a gait
analysis
tool
sports
and
medical
applications.
utilized
asas
a gait
analysis
tool
in in
sports
and
medical
applications.
Thesegmentation
segmentationwas
wascarried
carriedout
outbybya acustomized
customizedMATLAB
MATLABcode
codethat
thatdetects
detectsgroups
groupsofof
The
consecutivenon-zero
non‐zero
samples.Then
Thenit itsegments
segmentsthose
thosesignals
signalsinto
into
individualstance
stance
phases,
each
consecutive
samples.
individual
phases,
each
starting
with
a heel‐strike
and
ending
with
a toe
off.
Figure
shows
the
first
10‐m
trial
one
the
starting
with
a heel-strike
and
ending
with
a toe
off.
Figure
1818
shows
the
first
10-m
trial
ofof
one
ofof
the
participants, where four gait cycles were extracted after excluding the first and last two gait cycles.
Left foot vGRF was segmented into four stance phases (Figure 19A).
Sensors
2020,
957
Sensors
2020,
20,20,
957
2020
of of
2525
Sensors 2020, 20, 957
20 of 25
participants,where
wherefour
fourgait
gaitcycles
cycleswere
wereextracted
extractedafter
afterexcluding
excludingthe
thefirst
firstand
andlast
lasttwo
twogait
gaitcycles.
cycles.
participants,
Left
foot
vGRF
was
segmented
into
four
stance
phases
(Figure
19A).
Left
foot
vGRF
was
segmented
into
four
stance
phases
(Figure
19A).
(A)
(A)
(B)
(B)
Figure
(A)
Segmented
left‐foot
gait
cycles,
(B)
segmented
right‐foot
cycles.
Figure
19.
(A)
Segmented
left‐foot
gait
cycles,
(B)
segmented
right‐foot
cycles.
Figure
19.19.
(A)
Segmented
left-foot
gait
cycles,
(B)
segmented
right-foot
cycles.
vGRF [kg]
vGRF [kg]
The
data
were
sampled
with
asampling
sampling
rate
samples/second,
where
each
segment
(stance
The
data
were
sampled
with
rate
ofof
6060
samples/second,
where
each
segment
(stance
The
data
were
sampled
with
aa sampling
rate
of
60
samples/second,
where
each
segment
(stance
phase)takes
takes
around
Therefore,each
eachsegment
segmentconsists
consistsof
around42
samples,which
which
were
then
phase)
takes
around
Therefore,
each
segment
consists
ofof
around
4242
samples,
which
were
then
phase)
around
0.70.7s. s.
Therefore,
around
samples,
were
then
resampled
into
512
samples.
The
mean
values
and
standard
deviations
of
each
of
the
512
samples
resampled into
and
standard
deviations
of each
of the
with
resampled
into 512
512samples.
samples.The
Themean
meanvalues
values
and
standard
deviations
of each
of 512
the samples
512 samples
withrespect
respect
the4 4segmented
segmented
signals
werecalculated.
calculated.
Thenthe
themean
mean
values
along
withthe
the
respect
to theto
4to
segmented
signals signals
were
calculated.
Then theThen
mean
values
along
with
the deviation
with
the
were
values
along
with
deviation
from
the
means
(means
plus
and
minus
the
deviation)
the
left
foot
was
calculated
and
from
the means
(means
plus
and plus
minus
the
deviation)
for the left
foot
was
calculated
and plotted
deviation
from
the
means
(means
and
minus
the
deviation)
forfor
the
left
foot
was
calculated
and
plotted20).
(Figure
20).
Similar
steps
wererepeated
repeated
with
leftfoot
footvGRF
vGRF
(Figure
20).This
This
providesanan
(Figure
Similar
steps
were
repeated
with
left foot
vGRF
(Figure
20).(Figure
This provides
anprovides
illustrative
plotted
(Figure
20).
Similar
steps
were
with
left
20).
illustrative
figure
thatcan
canbe
usedbybyinsport
indifferent
different
sportand
and
medicalapplications
applications
asses
walking
figure
that figure
can
bethat
used
by
inbeused
different
and medical
applications
to asses walking
behaviors
illustrative
sport
medical
totoasses
walking
complications.
orbehaviors
complications.
behaviors
oror
complications.
Figure
Means
and
standard
deviations
of
gait
cycles;
blue
curves
represents
the
mean
gait
value
Figure
20.20.
Means
and
deviations
ofgait
gait
cycles;
blue
curves
represents
the
mean
gait
value
Figure
Means
standard
deviations
of
cycles;
blue
curves
represents
the
mean
gait
value
of
ofthe
theleft
leftfoot
footwith
withdashed
dashed
linerepresenting
representing
thedeviation
deviation
fromthe
the
mean,
while
orange
curves
of
line
the
mean,
while
orange
curves
the
left
foot
with
dashed
line representing
the deviation
from thefrom
mean,
while
orange
curves
represents
represents
the
mean
gait
value
of
the
right
foot
with
dashed
line
representing
the
deviation
from
the
represents
the
mean
gait
value
of
the
right
foot
with
dashed
line
representing
deviation
from
the
the
mean gait
value
of
the
right
foot
with
dashed
line
representing
the
deviationthe
from
the mean
value.
mean
value.
mean value.
The vGRF of a subject mainly depends on his/her health condition and the footwear used. In
this study,
all participants
were
advised
to wear
comfortable
walking
shoes
avoiding
high-heel
shoes,
The
vGRF
a subject
mainly
depends
his/her
health
condition
and
the
footwear
used.
In
this
The
vGRF
ofof
a subject
mainly
depends
onon
his/her
health
condition
and
the
footwear
used.
In
this
especially
for
female subjects.
This ensured
that
all subjects
went shoes
through
similar high‐heel
condition
while
study,allallparticipants
participants
wereadvised
advised
wear
comfortable
walking
shoesavoiding
avoiding
high‐heelshoes,
shoes,
study,
were
totowear
comfortable
walking
conducting
the
experiment.
ItThis
was
observed
that
the
collected
data
did not
showcondition
any
significant
especiallyforfor
female
subjects.
Thisensured
ensuredthat
that
subjects
went
through
similar
condition
while
especially
female
subjects.
allall
subjects
went
through
similar
while
statistical difference based on gender.
Sensors 2020, 20, 957
21 of 25
Sensors 2020, 20, 957
21 of 25
conducting the experiment. It was observed that the collected data did not show any significant
statistical difference based on gender.
5.2.2. Piezoelectric Sensor Based-Insole Characterization
5.2.2. Piezoelectric Sensor Based‐Insole Characterization
Three subjects participated in the piezo-electric insole test in the same manner as the testing of
Three subjects participated in the piezo‐electric insole test in the same manner as the testing of
FSR based insole. The piezoelectric insoles were expected to detect the gait cycle, with impulse signals
FSR based insole. The piezoelectric insoles were expected to detect the gait cycle, with impulse signals
in heel-area sensors during the heel strike phase and lower amplitude impulses from all the sensors
in heel‐area sensors during the heel strike phase and lower amplitude impulses from all the sensors
during the mid-stance phase. Finally, impulse signals from the toe and metatarsal heads sensors were
during the mid‐stance phase. Finally, impulse signals from the toe and metatarsal heads sensors were
taken in the toe-off phase. However, the readings were not promising, showing single irregular shape
taken in the toe‐off phase. However, the readings were not promising, showing single irregular shape
pulses per sensor for each individual gait cycle. The addition of different sensors output showed
pulses per sensor for each individual gait cycle. The addition of different sensors output showed
periodic impulses, one impulse per period (Figure 21). This indicates that the full stance period was
periodic impulses, one impulse per period (Figure 21). This indicates that the full stance period was
detected as one event only. Meanwhile, the correct vGRF must show two distinct peaks between
detected as one event only. Meanwhile, the correct vGRF must show two distinct peaks between the
the mid-stance phase, summing up to three main phases: heel strike, mid stance and toe off. The
mid‐stance phase, summing up to three main phases: heel strike, mid stance and toe off. The rigid
rigid nature of the piezo sensor made it difficult to detect different gait phases. Therefore, it can be
nature of the piezo sensor made it difficult to detect different gait phases. Therefore, it can be
summarized that it is not suitable for the smart insole application which requires to produce reliable
summarized that it is not suitable for the smart insole application which requires to produce reliable
vGRF signal due to gait.
vGRF signal due to gait.
Gait Cycles (Subject A)
100
Left Foot
Right Foot
vGRF [kg]
50
0
-50
0
1
2
3
4
Time [sec]
(a)
5
6
7
(b)
Figure21.
21.Gait
Gaitcycles
cyclesfor
forleft
leftand
andright
rightfoot
footwith
withpiezoelectric
piezoelectricsmart
smartinsole
insole(a)(A)
subject
1, (B)
subject
Figure
subject
1, (b)
subject
2.
2.
In this study, the authors have characterized three samples from each sensor category randomly;
however,
thestudy,
smartthe
insole
was have
implemented
usingthree
16-sensors.
was expected
there
In this
authors
characterized
samplesTherefore,
from each it
sensor
category that
randomly;
would
be a the
small
variation
the implemented
vGRF recordedusing
from 16‐sensors.
the smart insole
in different
trials
and inthat
different
however,
smart
insoleofwas
Therefore,
it was
expected
there
subjects.
However,
20 clearly
that the vGRF
individual
has a unique
would be
a small Figure
variation
of thedepicts
vGRF recorded
fromfrom
the an
smart
insole foot
in different
trialspattern
and in
and
this finding
matches
with the
vGRF
by the commercial
smart
insole
and force plate.
Thisa
different
subjects.
However,
Figure
20recorded
clearly depicts
that the vGRF
from
an individual
foot has
reflects
fact that
insole
designed
FSR is
capable by
of acquiring
vGRF smart
reliably
and the
uniquethe
pattern
andthe
thissmart
finding
matches
withusing
the vGRF
recorded
the commercial
insole
and
designed
system
robust enough
adapt
the age,
gender
and body
mass
(BMI) of
variation
of
force plate.
Thisisreflects
the fact to
that
the to
smart
insole
designed
using
FSRindex
is capable
acquiring
the
participants.
However,
carbon
piezoresistive
material
(like
Velostat),
which
authors
have
tested
in
vGRF reliably and the designed system is robust enough to adapt to the age, gender and body mass
preliminary
here However,
in order to carbon
avoid unnecessary
length
of the(like
manuscript),
index (BMI)experiments
variation of(not
the reported
participants.
piezoresistive
material
Velostat),
showed
highhave
hysteresis
type of material
is not suitable
for humanhere
dynamicity
monitoring.
which very
authors
testedand
inthis
preliminary
experiments
(not reported
in order
to avoid
On
the other hand,
sensors showed
can monitor
pressure
variation
theyisare
unnecessary
lengthpiezoelectric
of the manuscript),
very dynamic
high hysteresis
and
this typehowever,
of material
not
very
sensitive
to smalldynamicity
pressure change
and incapable
reliably
produce
mean vGRF.
Moreover,
the
suitable
for human
monitoring.
On the to
other
hand,
piezoelectric
sensors
can monitor
vGRF
changed
overvariation
trials and
over subjects
significantly
and,
temporal
feature
of vGRF
dynamic
pressure
however,
they are
very sensitive
totherefore,
small pressure
change
and incapable
cannot
be identified
piezoelectric
sensor-based
to reliably
produceusing
mean
vGRF. Moreover,
the smart
vGRF insole.
changed over trials and over subjects
significantly and, therefore, temporal feature of vGRF cannot be identified using piezoelectric sensor‐
5.3.
Performance
Evaluation of the FSR-Based System
based
smart insole.
Comparing the mean and standard deviation of vGRF for a gait cycle of the same subject recorded
5.3.
Performance
Evaluation
of the FSR‐Based
using the two systems,
commercial
F-scan System
and the proposed FSR insoles, it can be seen that both
showed
good
quality
signals
except
slight
differencesofinvGRF
peak for
values
(Figure
Thesame
FSR subject
insole
Comparing the mean and standard deviation
a gait
cycle 22).
of the
showed
smaller
vGRF
during
left-foot
heel
strike
phase
compared
to
the
F-scan
insole.
This
isthat
an
recorded using the two systems, commercial F‐scan and the proposed FSR insoles, it can be seen
expected behavior as each sensor is somewhat unique due to the manufacturing process and we cannot
Sensors 2020, 20, 957
22 of 25
both showed good quality signals except slight differences in peak values (Figure 22). The FSR insole
showed
smaller
vGRF during left‐foot heel strike phase compared to the F‐scan insole. This
Sensors
2020, 20,
957
22 ofis25an
expected behavior as each sensor is somewhat unique due to the manufacturing process and we
cannot calibrate individual sensors, which can lead to some variation. In addition, due to the presence
calibrate
individual
sensors, of
which
lead
to somemore
variation.
In addition,
dueshoe.
to the
of an insole,
the sensitivity
somecan
FSRs
decreases
than the
others in the
Topresence
mitigate of
this
anproblem,
insole, the
sensitivity
of
some
FSRs
decreases
more
than
the
others
in
the
shoe.
To
mitigate
this
a highly uniform pressure should be applied across individual sensors. Each sensor should
problem,
uniform
pressure
across
sensor should
producea highly
uniform
output.
When should
this is be
notapplied
the case
for individual
a specific sensors.
sensor, Each
the software
should
produce
uniform
output.
When
this
is
not
the
case
for
a
specific
sensor,
the
software
should
determine
determine a unique scale factor to compensate for the output variation. Currently, there are a few
a unique
scalesuch
factorastoT‐scan,
compensate
for the output
variation.
Currently,
there
are a few companies
companies
that provide
a special
piece of
equipment
(equilibration
device) such
which
asapplies
T-scan, athat
provide
a
special
piece
of
equipment
(equilibration
device)
which
applies
a
uniform
uniform pressure on the full insole using a thin flexible membrane to perform such
pressure
on the
full insole
using a to
thin
perform
such
calibration.
calibration.
Moreover,
compared
theflexible
F‐scan membrane
system, theto
FSR
readings
showed
smallerMoreover,
differences
compared
the F-scan
system,
FSRvalues.
readings
smaller
between
vGRF
between to
vGRF
peaks and
mid the
stance
Thisshowed
was mainly
duedifferences
to the superior
number
ofpeaks
sensors
and
mid
stance
values.
This
was
mainly
due
to
the
superior
number
of
sensors
for
the
F-scan
system
for the F‐scan system (960 sensing areas) compared to the proposed insole (16 FSRs). In addition, the
(960
sensing
areas)elements
compared
to the
proposed
insole (16 on
FSRs).
In addition,
the
F-scan
elements
F‐scan
sensing
were
uniformly
distributed
the full
foot area,
while
thesensing
FSR sensors
were
were
uniformly
on the most
full foot
area,
while the
FSR sensors
placedplaced
on the foot
areas
placed
on the distributed
foot areas where
of the
pressure
is exerted
withwere
no sensors
on the
low‐
where
most
of
the
pressure
is
exerted
with
no
sensors
placed
on
the
low-pressure
areas
(medial
arch).
pressure areas (medial arch). Adding a few FSR sensors to the medial arch can improve the quality
Adding
few FSR
sensors to
the medial
can improve
thefoot,
quality
of the
signal
obtained, especially
of theasignal
obtained,
especially
forarch
subjects
with flat
who
exerts
considerable
amount of
forpressure
subjectson
with
flat
foot,
who
exerts
considerable
amount
of
pressure
on
medial
arch
areas.
medial arch areas.
A
B
Figure
Comparison
between
mean
and
standard
deviation
vertical
ground
reaction
forces
Figure
22.22.
Comparison
between
thethe
mean
and
standard
deviation
of of
vertical
ground
reaction
forces
(vGRF)
from
(blue)
and
right
(orange)
foot
using
F‐scan
system
and
FSR‐system
(vGRF)
from
leftleft
(blue)
and
right
(orange)
foot
using
F-scan
system
(A)(A)
and
FSR-system
(B).(B).
6. 6.
Conclusions
Conclusion
InIn
this
study,
thethe
authors
have
proposed
and
designed
low-cost
calibration
setups
forfor
calibrating
this
study,
authors
have
proposed
and
designed
low‐cost
calibration
setups
calibrating
three
different
force
sensors:
FSR,
ceramic
piezoelectric
and
flexible
piezoelectric
sensors.
three different force sensors: FSR, ceramic piezoelectric and flexible piezoelectric sensors.The
The
experiments
experimentsconducted
conductedshowed
showedthe
theeffectiveness
effectivenessofofthe
theproposed
proposedsetup
setupinincalibrating
calibratingFSR
FSRand
and
piezoelectric
sensors,
which
are mainly
affectedaffected
by 1D force.
It was
found
flexible
piezoelectric
piezoelectric
sensors,
which
are mainly
by 1D
force.
It that
was the
found
that
the flexible
sensors
were
performing
poor
in
terms
of
calibration
due
to
their
sensitivity
to
3D
forces.
Special
piezoelectric sensors were performing poor in terms of calibration due to their sensitivity to 3D
forces.
force
calibration
machines
are
required
to
control
the
applied
force
in
x,
y
or
z
directions.
In
addition,
Special force calibration machines are required to control the applied force in x, y or z directions. In
a systematic
designing
characterizing
two differenttwo
smart
insolessmart
were insoles
illustrated.
addition, aprocedure
systematicfor
procedure
forand
designing
and characterizing
different
were
The
vGRF
signal
acquired
and
segmented
to
obtain
mean
vGRF
and
its
standard
deviation
for
a
gait
illustrated. The vGRF signal acquired and segmented to obtain mean vGRF and its standard deviation
cycle
can be used
measure
different
statistical
metricsstatistical
(such as mean
standard
for were
a gaitcalculated,
cycle werewhich
calculated,
whichtocan
be used
to measure
different
metrics
(such as
deviation,
time
to
peak,
etc.)
that
can
help
in
assessing
the
walking
behavior
of
athletes,
patients
orof
mean standard deviation, time to peak, etc.) that can help in assessing the walking behavior
normal
people.
The
FSR-based
smart
insole
was
able
to
acquire
high
quality
vGRF
for
different
gait
athletes, patients or normal people. The FSR‐based smart insole was able to acquire high quality
cycles.
other hand,
the piezoelectric
sensor-based
insole failedsensor‐based
to detect distinct
phases.
It
vGRFOn
forthe
different
gait cycles.
On the other
hand, the piezoelectric
insolegait
failed
to detect
cannot
be
utilized
as
an
alternative
to
FSR
in
smart
insole
application.
However,
the
calibrated
piezo
distinct gait phases. It cannot be utilized as an alternative to FSR in smart insole application.
sensors
can be
in other
bio-sensing
technologies
as detecting
the start
and end ofsuch
eachas
However,
theutilized
calibrated
piezo
sensors can
be utilizedsuch
in other
bio‐sensing
technologies
gait
cycle. the start and end of each gait cycle.
detecting
Sensors 2020, 20, 957
23 of 25
Author Contributions: Experiments were designed by M.E.H.C., N.E. and A.K. Experiments were performed
and experimental data were acquired by A.M.T., S.A.-H., M.A., and S.A. Data were analyzed by A.M.T., M.E.H.C.,
A.K. and N.A.-E. All authors were involved in interpretation of data and writing the paper. All authors have read
and agreed to the published version of the manuscript.
Funding: This research was partially funded by Qatar National Research Foundation (QNRF), grant number
NPRP12S-0227-190164 and Research University Grant DIP-2018-017. The publication of this article was funded by
the Qatar National Library.
Acknowledgments: The authors would like to thank Engr. Ayman Ammar, Electrical Engineering, Qatar
University for helping in printing the printed circuit boards (PCBs).
Conflicts of Interest: The authors declare no conflict of interest.
References
1.
2.
3.
4.
5.
6.
7.
8.
9.
10.
11.
12.
13.
14.
15.
Kowalski, E.; Catelli, D.S.; Lamontagne, M. Side does not matter in healthy young and older
individuals–Examining the importance of how we match limbs during gait studies. Gait Posture 2019,
67, 133–136.
Tao, W.; Liu, T.; Zheng, R.; Feng, H. Gait analysis using wearable sensors. Sensors 2012, 12, 2255–2283.
[CrossRef] [PubMed]
Howell, A.M.; Kobayashi, T.; Hayes, H.A.; Foreman, K.B.; Bamberg, S.J.M. Kinetic gait analysis using a
low-cost insole. IEEE Trans. Biomed. Eng. 2013, 60, 3284–3290. [CrossRef] [PubMed]
Bamberg, S.J.M.; Benbasat, A.Y.; Scarborough, D.M.; Krebs, D.E.; Paradiso, J.A. Gait analysis using a
shoe-integrated wireless sensor system. IEEE Trans. Inf. Technol. Biomed. 2008, 12, 413–423. [CrossRef]
[PubMed]
De Rossi, S.; Lenzi, T.; Vitiello, N.; Donati, M.; Persichetti, A.; Giovacchini, F.; Vecchi, F.; Carrozza, M.C.
Development of an in-shoe pressure-sensitive device for gait analysis. In Proceedings of the 2011 Annual
International Conference of the IEEE Engineering in Medicine and Biology Society, Boston, MA, USA,
30 August–3 September 2011; IEEE: Piscataway, NJ, USA, 2011.
Smart Tracking and Wearables: Techniques in Gait Analysis and Movement in Pathological Aging. Available
online: https://www.intechopen.com/books/smart-healthcare/smart-tracking-and-wearables-techniquesin-gait-analysis-and-movement-in-pathological-aging (accessed on 9 February 2020).
Gurchiek, R.D.; Choquette, R.H.; Beynnon, B.D.; Slauterbeck, J.R.; Tourville, T.W.; Toth, M.J.; McGinnis, R.S.
Remote gait analysis using wearable sensors detects asymmetric gait patterns in patients recovering from
ACL reconstruction. In Proceedings of the 2019 IEEE 16th International Conference on Wearable and
Implantable Body Sensor Networks (BSN), Chicago, IL, USA, 19–22 May 2019; IEEE: Piscataway, NJ, USA,
2019.
Brognara, L.; Palumbo, P.; Grimm, B.; Palmerini, L. Assessing gait in parkinson’s disease using wearable
motion sensors: A systematic review. Diseases 2019, 7, 18. [CrossRef] [PubMed]
Mills, P.M.; Morrison, S.; Lloyd, D.G.; Barrett, R.S. Repeatability of 3D gait kinematics obtained from an
electromagnetic tracking system during treadmill locomotion. J. Biomech. 2007, 40, 1504–1511. [CrossRef]
Huang, E.; Sharp, M.T.; Osborn, E.; MacLellan, A.; Mlynash, M.; Kemp, S.; Buckwalter, M.S.; Lansberg, M.G.
Abstract TP173: Feasibility and Utility of Home-Based Gait Analysis Using Body-Worn Sensors. Stroke 2019,
50, ATP173. [CrossRef]
Liu, T.; Inoue, Y.; Shibata, K. A wearable ground reaction force sensor system and its application to the
measurement of extrinsic gait variability. Sensors 2010, 10, 10240–10255. [CrossRef] [PubMed]
Razak, A.; Hadi, A.; Zayegh, A.; Begg, R.K.; Wahab, Y. Foot plantar pressure measurement system: A review.
Sensors 2012, 12, 9884–9912. [CrossRef] [PubMed]
Alam, U.; Riley, D.R.; Jugdey, R.S.; Azmi, S.; Rajbhandari, S.; D’Août, K.; Malik, R.A. Diabetic neuropathy
and gait: A review. Diabetes Ther. 2017, 8, 1253–1264. [CrossRef] [PubMed]
Farahpour, N.; Jafarnezhad, A.; Damavandi, M.; Bakhtiari, A.; Allard, P. Gait ground reaction force
characteristics of low back pain patients with pronated foot and able-bodied individuals with and without
foot pronation. J. Biomech. 2016, 49, 1705–1710. [CrossRef] [PubMed]
Sacco, I.C.; Picon, A.P.; Macedo, D.O.; Butugan, M.K.; Watari, R.; Sartor, C.D. Alterations in the lower limb
joint moments precede the peripheral neuropathy diagnosis in diabetes patients. Diabetes Technol. Ther. 2015,
17, 405–412.
Sensors 2020, 20, 957
16.
17.
18.
19.
20.
21.
22.
23.
24.
25.
26.
27.
28.
29.
30.
31.
32.
33.
34.
35.
24 of 25
Dabbah, M.A.; Graham, J.; Petropoulos, I.N.; Tavakoli, M.; Malik, R.A. Automatic analysis of diabetic
peripheral neuropathy using multi-scale quantitative morphology of nerve fibres in corneal confocal
microscopy imaging. Med. Image Anal. 2011, 15, 738–747. [CrossRef] [PubMed]
International Diabetes Federation: IDF Diabetes Atlas. Available online: https://www.idf.org/e-library/
epidemiology-research/diabetes-atlas/19-atlas-6th-edition.html (accessed on 9 February 2020).
Dworkin, R.H.; Malone, D.C.; Panarites, C.J.; Armstrong, E.P.; Pham, S.V. Impact of postherpetic neuralgia
and painful diabetic peripheral neuropathy on health care costs. J. Pain 2010, 11, 360–368. [CrossRef]
Dyck, P.J.; Overland, C.J.; Low, P.A.; Litchy, W.J.; Davies, J.L.; Dyck, P.J.B.; O’brien, P.C. Signs and symptoms
versus nerve conduction studies to diagnose diabetic sensorimotor polyneuropathy: Cl vs. NPhys trial.
Muscle Nerve 2010, 42, 157–164.
Taksande, B.; Ansari, S.; Jaikishan, A.; Karwasara, V. The diagnostic sensitivity, specificity and reproducibility
of the clinical physical examination signs in patients of diabetes mellitus for making diagnosis of peripheral
neuropathy. J. Endocrinol. Metab. 2011, 1, 21–26. [CrossRef]
Watari, R.; Sartor, C.D.; Picon, A.P.; Butugan, M.K.; Amorim, C.F.; Ortega, N.R.; Sacco, I.C. Effect of diabetic
neuropathy severity classified by a fuzzy model in muscle dynamics during gait. J. Neuroeng. Rehabil. 2014,
11, 11. [CrossRef]
England, J.; Gronseth, G.; Franklin, G.; Carter, G.; Kinsella, L.; Cohen, J.; Asbury, A.; Szigeti, K.; Lupski, J.;
Latov, N. Practice Parameter: Evaluation of distal symmetric polyneuropathy: Role of laboratory and genetic
testing (an evidence-based review) Report of the American Academy of Neurology, American Association
of Neuromuscular and Electrodiagnostic Medicine, and American Academy of Physical Medicine and
Rehabilitation. Neurology 2009, 72, 185–192.
Zhang, Q.; Wang, Y.L.; Xia, Y.; Wu, X.; Kirk, T.V.; Chen, X.D. A low-cost and highly integrated sensing insole
for plantar pressure measurement. Sens. Bio-Sens. Res. 2019, 26, 100298. [CrossRef]
Zhang, W.; Yin, B.; Wang, J.; Mohamed, A.; Jia, H. Ultrasensitive and wearable strain sensors based on
natural rubber/graphene foam. J. Alloys Compd. 2019, 785, 1001–1008. [CrossRef]
Sengupta, D.; Pei, Y.; Kottapalli, A.G.P. Ultralightweight and 3D Squeezable Graphene-Polydimethylsiloxane
Composite Foams as Piezoresistive Sensors. ACS Appl. Mater. Interfaces 2019, 11, 35201–35211. [CrossRef]
[PubMed]
Shu, L.; Hua, T.; Wang, Y.; Li, Q.; Feng, D.D.; Tao, X. In-shoe plantar pressure measurement and analysis
system based on fabric pressure sensing array. IEEE Trans. Inf. Technol. Biomed. 2010, 14, 767–775. [PubMed]
Giovanelli, D.; Farella, E. Force sensing resistor and evaluation of technology for wearable body pressure
sensing. J. Sens. 2016, 2016. [CrossRef]
Feasibility of Force Detection in 3D Printed Flexible Material Using Embedded Sensors. Available
online:
https://www.spiedigitallibrary.org/conference-proceedings-of-spie/10970/109702F/Feasibilityof-force-detection-in-3D-printed-flexible-material-using/10.1117/12.2514051.short?SSO=1 (accessed on
9 February 2020).
Heng, W.; Pang, G.; Xu, F.; Huang, X.; Pang, Z.; Yang, G. Flexible Insole Sensors with Stably Connected
Electrodes for Gait Phase Detection. Sensors 2019, 19, 5197. [CrossRef]
Lee, W.; Hong, S.-H.; Oh, H.-W. Characterization of Elastic Polymer-Based Smart Insole and a Simple Foot
Plantar Pressure Visualization Method Using 16 Electrodes. Sensors 2019, 19, 44. [CrossRef]
Theodosiou, A.; Kalli, K. Recent trends and advances of fibre Bragg grating sensors in CYTOP polymer
optical fibres. Opt. Fiber Technol. 2020, 54, 102079. [CrossRef]
Abdelhady, M.; van den Bogert, A.; Simon, D. A High-Fidelity Wearable System for Measuring Lower-Limb
Kinetics and Kinematics. IEEE Sens. J. 2019. [CrossRef]
Kim, J.; Bae, M.N.; Lee, K.B.; Hong, S.G. Gait event detection algorithm based on smart insoles. ETRI J. 2019.
[CrossRef]
Vilas-Boas, M.d.C.; Rocha, A.P.; Choupina, H.M.P.; Cardoso, M.; Fernandes, J.M.; Coelho, T.; Cunha, J.P.S.
TTR-FAP Progression Evaluation Based on Gait Analysis Using a Single RGB-D Camera. In Proceedings of
the 2019 41st Annual International Conference of the IEEE Engineering in Medicine and Biology Society
(EMBC), Berlin, Germany, 23–27 July 2019; IEEE: Piscataway, NJ, USA, 2019.
Cui, X.; Zhao, Z.; Ma, C.; Chen, F.; Liao, H. A Gait Character Analyzing System for Osteoarthritis Pre-diagnosis
Using RGB-D Camera and Supervised Classifier. In Proceedings of the World Congress on Medical Physics
and Biomedical Engineering, Prague, Czech Republic, 3–8 June 2018; Springer: Berlin, Germany, 2018.
Sensors 2020, 20, 957
36.
37.
38.
39.
40.
41.
42.
43.
44.
45.
46.
47.
48.
49.
25 of 25
Barnea, A.; Oprisan, C.; Olaru, D. Force Sensitive Resistors Calibration for the Usage in Gripping Devices.
Diagn. Predict. Mech. Eng. Szstems DIPRE 2012, 3, 18–27.
Parmar, S.; Khodasevych, I.; Troynikov, O. Evaluation of flexible force sensors for pressure monitoring in
treatment of chronic venous disorders. Sensors 2017, 17, 1923. [CrossRef]
Insole Energy Harvesting from Human Movement Using Piezoelectric Generators. Available online:
https://trepo.tuni.fi/handle/123456789/25768 (accessed on 9 February 2020).
Development of a Smart Insole System for Real-Time Detection of Temporal Gait Parameters and Related
Deviations in Unilateral Lower-Limb Amputees. Available online: https://scholarlyrepository.miami.edu/
cgi/viewcontent.cgi?article=1393&context=oa_theses (accessed on 9 February 2020).
Sacco, I.; Amadio, A. A study of biomechanical parameters in gait analysis and sensitive cronaxie of diabetic
neuropathic patients. Clin. Biomech. 2000, 15, 196–202. [CrossRef]
Menz, H.B.; Lord, S.R.; St George, R.; Fitzpatrick, R.C. Walking stability and sensorimotor function in older
people with diabetic peripheral neuropathy1. Arch. Phys. Med. Rehabil. 2004, 85, 245–252. [CrossRef]
[PubMed]
Dementyev, A.; Hodges, S.; Taylor, S.; Smith, J. Power consumption analysis of Bluetooth Low Energy, ZigBee
and ANT sensor nodes in a cyclic sleep scenario. In Proceedings of the 2013 IEEE International Wireless
Symposium (IWS), Beijing, China, 14–18 April 2013; IEEE: Piscataway, NJ, USA, 2013.
Sacco, I.; Amadio, A. Influence of the diabetic neuropathy on the behavior of electromyographic and sensorial
responses in treadmill gait. Clin. Biomech. 2003, 18, 426–434. [CrossRef]
Kwon, O.-Y.; Minor, S.D.; Maluf, K.S.; Mueller, M.J. Comparison of muscle activity during walking in subjects
with and without diabetic neuropathy. Gait Posture 2003, 18, 105–113.
Akashi, P.M.; Sacco, I.C.; Watari, R.; Hennig, E. The effect of diabetic neuropathy and previous foot ulceration
in EMG and ground reaction forces during gait. Clin. Biomech. 2008, 23, 584–592. [CrossRef] [PubMed]
Sacco, I.C.; Akashi, P.M.; Hennig, E.M. A comparison of lower limb EMG and ground reaction forces between
barefoot and shod gait in participants with diabetic neuropathic and healthy controls. BMC Musculoskelet.
Disord. 2010, 11, 24. [CrossRef]
Gomes, A.A.; Onodera, A.N.; Otuzi, M.E.; Pripas, D.; Mezzarane, R.A.; Sacco, N.; Isabel, C. Electromyography
and kinematic changes of gait cycle at different cadences in diabetic neuropathic individuals. Muscle Nerve
2011, 44, 258–268.
Sawacha, Z.; Spolaor, F.; Guarneri, G.; Contessa, P.; Carraro, E.; Venturin, A.; Avogaro, A.; Cobelli, C.
Abnormal muscle activation during gait in diabetes patients with and without neuropathy. Gait Posture 2012,
35, 101–105.
Zhu, H.; Wertsch, J.J.; Harris, G.F.; Loftsgaarden, J.D.; Price, M.B. Foot pressure distribution during walking
and shuffling. Arch. Phys. Med. Rehabil. 1991, 72, 390–397.
© 2020 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/).