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

net/publication/329495752

Low-Cost Pressure Sensor Matrix Using Velostat

Conference Paper · November 2017

DOI: 10.1109/ICICI-BME.2017.8537720

CITATIONS READS

0 675

5 authors, including:

Sena Sukmananda Suprapto Hazem M Zakaria

Institut Teknologi Kalimantan Imam Abdul Rahman bin Faisal University
3 PUBLICATIONS   4 CITATIONS    48 PUBLICATIONS   307 CITATIONS   

SEE PROFILE SEE PROFILE

Widyawardana Adiprawita Basuki Supartono

Bandung Institute of Technology air force hospital of indonesia
60 PUBLICATIONS   228 CITATIONS    9 PUBLICATIONS   2 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Wearable Fall Detection View project

Auditory Neuroscience View project

All content following this page was uploaded by Sena Sukmananda Suprapto on 16 July 2019.

The user has requested enhancement of the downloaded file.

 2017 5th International Conference on Instrumentation, Communications, Information Technology, and Biomedical Engineering (ICICI-BME)
Bandung, 6-7 November 2017

Low-Cost Pressure Sensor Matrix Using Velostat

S.S. Suprapto1, A.W. Setiawan2, H. Zakaria3, W. Adiprawita4 B. Supartono

Sekolah Teknik Elektro dan Informatika, Medical Faculty,
Institut Teknologi Bandung, UPN Veteran Jakarta,
Bandung, Indonesia Jakarta, Indonesia
[email protected], [email protected], [email protected]
[email protected], [email protected]

Abstract— This paper presents a characteristic of a 32x32 One of the most widely used sensors in the application of
sensor matrix based on velostat for foot pressure distribution plantar pressure is the force sensing resistor (FSR). Force
measurements. The sensor has been constructed of dual Sensing Resistor (FSR) is an analog sensor that function to
electrodes placed on the top and underneath velostat as the change the compression force to change the resistance, when
sensing material. The experimental results show the sensor given the force on the FSR, the resistance will decrease. FSRs
matrix has relatively linear in Force-Conductance response until are made of a conductive polymer that changes resistance with
15 N. The system evaluation stage was obtained by comparing the force, applying force causes conductive particles to touch
sensor matrix with several other methods implemented in RSON increasing the current through the sensors. The advantages of
and obtained the performance of the sensor matrix in the form of
FSR are high spatial resolution and flexible where FlexiForce
the comparison of the footprint area with the error value of 7.3%
A301 has range from 10-110.000 N with 0.5 inch diameter[5].
to 68.8%, the average pressure of 7.3% to 68.8%, the maximum
pressure 37.6% to 70.7%, body weight of 0.4% to 12.1%, and
The advantages besides having high spatial resolution and
arch-index with a specificity value of 0.3 and a sensitivity of 0.5. flexible, the resistive sensor is also low-cost. Many researchers
have conducted measurements of pressure and posture analysis
Keywords—Velostat; Pressure Sensor Matrix; Crosstalk using the FSR sensor design results, specification and signal
conditioning designs according to application needs.
I. INTRODUCTION Del Prete et al.[6] pressure measurement in analyzing in the
The distribution of human plantar pressure data reflect nature of metrological for static and dynamic measurement
conditions associated with the foot, such as foot structure, conditions, the result of design using an 8×8 matrix array of
function and control the posture of the whole body and others. pressure sensors Novel (velostat) and the result of pressure
Human foot pressure distribution measurement can provide from the system design of 0-5000 kPa. D. Giovanelli et al.[2]
vital information and very useful for medical diagnosis. One implemented in a wearable technology, using FSR sensor with
mayor that has attracted considerable attention by researchers the resistive material (velostat) for pressure measurement. Lin
in biomedical and sport related applications is an analysis of and Seet [7] using a 10 mm×10 mm textile pressure sensor to
foot plantar pressure distribution to reveal the interface be implemented in pressure measurement with 0-1000 kPa
pressure between the foot plantar surface and shoe sole. pressure design results. This paper presents a characterized by
velostat sensors, implementation, and evaluation of appropriate
The most common pressure sensors are capacitive sensors, system platforms for foot pressure distribution measurements.
resistive sensors, piezoelectric sensor and piezoresistive sensor.
Capacitive force sensors are usually a parallel-plate capacitor, II. SENSOR DESIGN
which changes its capacitance in function of the applied force.
This is due to the dielectric that, in this case, is an another sort The platform system is designed with a 32×32 matrix and
of squeezable material that gets thinner when subject to each sensor size is 7mm×7mm with a distance between sensors
pressure[1]. The advantages of capacitive sensors are high is 1cm to avoid a short circuit between sensors. The sensor is
accuracy, Flexible, durable. However, the conditioning circuit constructed of three main layers which are a top electrode,
and computation are complexes[2]. Piezoelectric force sensors velostat, and bottom electrode respectively.
are based on the piezoelectric effect of some materials, which Sensor matrix is controlled by the microcontroller to scan
generate an electric charge when stressed. The most suitable every single sensor one by one using multiplexer.
material for clinically oriented body pressure measurement is Unfortunately, this setup is causing leak current to others
polyvinylidene fluoride (PVDF). These flexible materials can sensor while scanning specific sensor (cross-talk). It's mean
be used to sense force or pressure but it is sensitive to many others sensor value could affect the scanned sensor.
factors (i.e., bending, temperature)[3]. Piezoresistive/force
sensing resistors instead have the advantages that can be To avoid cross-talk, the current barrier is applied to every
fabricated using flexible materials, but also they are very robust sensor. To achieve this setup, double layer PCB is used as a
against noise and the conditioning electronics is simple, that in non-conductive layer to connect the top electrode to a diode
many cases only a bias resistor is used[4]. which can simplified manufacturing process.

978-1-5386-3455-4/17/$31.00 ©2017 IEEE 137

2017 5th International Conference on Instrumentation, Communications, Information Technology, and Biomedical Engineering (ICICI-BME)
Bandung, 6-7 November 2017

1
G( velostat ) = (2)
R( velostat )

Where G(velostat)=sensor matrix conductance.

Figure 4 shows the wide deviation of the sensor sample.
The pattern has been searched for the cause of the sample
sensor wide deviation value. But no pattern can be represented
by the random form of the deviation. Therefore this wide
deviation value is possible because it is the influence of hand-
made manufacturing.
Fig. 1. Three main layers of sensor matrix
Although deviation of 64 samples on sensor matrix is wide,
it percentage tends to reduce as the force increases. This
evidance has been shown in Figure 5 which illustrates the
relationship between the two. Its mean, this hand-made sensor
matrix more realible to sense high force than low force.
Measurements show that the sensor matrix has most likely
power-equation relation in Force-Resistance and relatively
linear-equation in Force-Conductance response until 15 N,
which can be modeled by the following equations:
1
Fig. 2. Current barrier to avoid cross-talk
§ 6689.68 · 0.88
F =¨ ¸ (3)
III. CALIBRATION © R( velostat ) ¹
Sensor calibration is done by measuring force to output
voltage response. To get an overview of the system response, F = 8879.18G( velostat ) − 0.76 (4)
use samples as 8x8 (64 samples) with interval 4 sensors in each
row and column. This sampling shortens the working time yet Where R(velostat)=sensor matrix resistance,
still get the representative of the system response from each G(velostat)=sensor matrix conductance and F=applied weight
part of the sensor matrix. force.

The instrument for calibration is a balance weight set which From the equations, we get the Square Error value of
has range 1 to 15 N with interval 1 N. In this experiment the equation 3 and equation 4 is 1.84 and 0.93 respectively. Based
sensor matrix has been placed on a flat surface. Since this on this, it is found that the relationship between conductance
matrix is hand-made and the electrode has a relatively small and compressive forces has a smaller Square Error value. Also
size, the sensor matrix thickness is not strictly uniformed. By in equation 4, the relationship between the two is linear where
placing the fabric (soft and non-conductive) as a cushion it is simpler compared to the relationship in equation 3.
between the balance weight set with the sensor matrix, the The median filter is used in this system to reduce the noise.
force on the sensor matrix can be spread evenly. But with this pressure with a small surface area becomes
Output value has been measured 30 seconds after load unreadable by the system. Given the area of the object to be
application to make sure voltage measurement in steady state. used has a size larger than the sensor cell size on the sensor
The sensor calibration curve has been measured by converting matrix then this is not a problem. By using the median filter we
10bit ADC value to resistance by the following equation: get the following matrix sensor outputs.

1024 − V( ADC )
R( velostat ) = R( reference ) (1)
V( ADC )

Where V(ADC)=10 bit ADC value, R(reference)=resistor

reference on voltage divider, and R(velostat)=sensor matrix
resistance.
The time response has been measured for a 15N step unit.
There is no response delay observe and the time response is
350-350ms as shown in figure 3.
The Force-Resistance and the Force-Conductance curve are
obtained by averaging the entire sample value (64 samples). Fig. 3. The sensor output response during a 15 N step loading
The Force-Conductance curve which sensor matrix
conductance can be calculated by the following equation:

138
 2017 5th International Conference on Instrumentation, Communications, Information Technology, and Biomedical Engineering (ICICI-BME)
Bandung, 6-7 November 2017

TABLE I. THREE OTHER MEASUREMENT METHODS TO EVALUATE

SENSOR MATRIX VELOSTAT

Average Total Body Max

Instrument Area Arch-index
Force Weight Force
Sensor Matrix Velostat ‫ط‬ ‫ط‬ ‫ط‬ ‫ط‬ ‫ط‬
Footprint (Ink and Paper) ‫ط‬ ‫ع‬ ‫ع‬ ‫ع‬ ‫ط‬
PEDOSCAN ‫ط‬ ‫ط‬ ‫ط‬ ‫ط‬ ‫ع‬
Scale ‫ع‬ ‫ع‬ ‫ط‬ ‫ع‬ ‫ع‬
Where,
‫ = ط‬Gold Standard Measurement
‫ = ط‬Tested
‫ = ع‬Untested

The test was conducted using 10 people who have weight

between 40 kg to 100 kg. The selection of body weight above
Fig. 4. Force-Resistive and Force-Conductance characteristic of sensor
matrix. Deviation is representation of 64 samples sensor 40 kg is done to get the high pressure value of the sole of the
foot to be read by the matrix sensor, while the selection of body
weight below 100 kg to keep the matrix sensor from the
possibility to be damaged. The measurement is done 3 times to
obtain a value close to the true value of the output of each
method. As for the measurement of ink and paper footprint
using only 1 time in accordance with the procedures in RSON.

Fig. 5. Force-Standard Deviation of Conductance characteristic. Deviation is

representation of 64 samples sensor

Fig. 6. Before and after median filter applied to the resuld of sensor matrix
velostat

IV. EXPERIMENTS AND RESULTS

Comparisons to other measurement methods were
performed to test the overall sensor matrix. The test is
performed in RSON (Rumah Sakit Olahraga Nasional) by
comparing the results from the matrix sensor with three other
methods which are Scale, PEDOSCAN, and a footprint
measurement using ink and paper. The compared outputs are
maximum pressure, average pressure, surface area and arch-
index. The output of each method are as follows: Fig. 7. The comparation of the result (area, average pressure, total body
weight, maximum pressure, and arch-index) from four measurement methods

139
2017 5th International Conference on Instrumentation, Communications, Information Technology, and Biomedical Engineering (ICICI-BME)
Bandung, 6-7 November 2017

V. DISCUSSION Respondent 4 Normal Normal FlatFeet FlatFeet F(-) F(-)

Respondent 5 Normal Normal Normal Normal T(+) T(+)
The footprint surface area is the contact area of foot while Respondent 6 FlatFeet FlatFeet FlatFeet FlatFeet T(-) T(-)
in standing position. From Fig.7, it can be seen that the surface Respondent 7 HighArch HighArch Normal Normal F(+) F(+)
Respondent 8 Normal Normal Normal HighArch T(+) F(-)
area using PEDOSCAN is much higher than the footprint Respondent 9 HighArch Normal FlatFeet HighArch F(-) F(-)
method. The difference that occurs is 61.7% to 158.6%. this Respondent 10 Normal Normal Normal Normal T(+) T(+)
happened because the upper surface of PEDOSCAN has a Based on table asdf, specificity and sensitivity can be
foam-like property. It is possible this large difference value calculated with this formula:
because the footprint method is done on a flat and rigid surface
(floor), whereas in PEDOSCAN the foot sinks on the surface T ( −)
Specificity = (5)
of the foam causing contact between the PEDOSCAN with the
T ( −) + F ( + )
foot to become wider. The sensor matrix is also the measured
higher than the footprint method. However, the value of the Specificity = 0.3 (6)
difference is not as high as PEDOSCAN is 7.3% to 68.8%.
The average pressure is the average pressure of the foot. By
comparing the mean pressure of the PEDOSCAN with the T ( +)
matrix sensor then seen in Fig.7 the measurement have Sensitivity = (7)
consistency of both methods. Where when the PEDOSCAN T ( +) + F (−)
shows the average left foot pressure value greater than the right
foot, the matrix sensor also shows the same thing and when the Sensitivity = 0.5 (8)
mean left foot pressure value is less than the right foot, the
matrix sensor also shows the same thing. VI. CONCLUSION
The interesting thing has shown in the weight-loss Although deviation of 64 samples on sensor matrix is wide,
measurement, where its value is related to the average footing measurements show that the sensor matrix has relatively linear-
and pressure. By using the weight scale we get the weight of equation in Force-Conductance response until 15 N and has
each respondent which is then compared with the output of the Square Error Value=0.93: By comparing the sensor matrix
PEDOSCAN and the sensor matrix. The output of velostat with other methods, the result of sensor matrix is
PEDOSCAN has a value above the weight scale with a closer to weight scale than PEDOSCAN. However, sensor
difference of 30.9% to 60.6%. while the sensor matrix only has matrix velostat still cannot replace footprint method for
a difference of 0.4% to 12.1%. This shows that the calculating arch-index due to sensitivity and specificity are too
PEDOSCAN in RSON can not be used as a reference in taking low.
weight data because the value is too far compared to weight
scale as gold standard weight measurement. REFERENCES
The maximum pressure is the highest pressure of foot. In [1] L. Viry et al., “Flexible Three-Axial Force Sensor for Soft and Highly
Fig.7, the result difference of the two methods is very far, Sensitive Artificial Touch,” Adv. Mater., vol. 26, no. 17, pp. 2659–2664,
where the PEDOSCAN can reach 29.7 N/cm2 but the sensor May 2014.
matrix only reaches 8.9 N/cm2. This might be caused by [2] D. Giovanelli and E. Farella, “Force Sensing Resistor and Evaluation of
Technology for Wearable Body Pressure Sensing,” J. Sens., vol. 2016,
imprecise reading of PEDOCSAN. But because the result 2016.
difference of the two methods is too far, it may be due to other
[3] L. Seminara, M. Capurro, P. Cirillo, G. Cannata, and M. Valle,
factors. “Electromechanical characterization of piezoelectric PVDF polymer
films for tactile sensors in robotics applications,” Sens. Actuators Phys.,
Arch-index is the index of curvature of the foot obtained by vol. 169, no. 1, pp. 49–58, Sep. 2011.
calculating the area of 1/3 the middle of the foot divided by the
[4] K. Flores De Jesus, M. H. Cheng, L. Jiang, and E. G. Bakhoum,
surface area where the area of the toe is not included in the “Resolution Enhancement Method Used for Force Sensing Resistor
calculation which is divided into three categories, high arch for Array,” J. Sens., vol. 2015, pp. 1–12, 2015.
arch-index <0.21, normal for arch-index 0.21 to 0.26, and flat [5] B. Wang, K. S. Rajput, W.-K. Tam, A. K. Tung, and Z. Yang,
foot for arch-index> 0.26. “FreeWalker: A smart insole for longitudinal gait analysis,” in
Engineering in Medicine and Biology Society (EMBC), 2015 37th
Annual International Conference of the IEEE, 2015, pp. 3723–3726.
TABLE II. THE ARCH INDEX TRUE CONDITION FROM SENSOR MATRIX [6] Z. Del Prete, L. Monteleone, and R. Steindler, “A novel pressure array
VELOSTAT CALCULATION COMPARED TO FOOTPRINT METHOD sensor based on contact resistance variation: Metrological properties,”
Respondent Footprint Sensor Matrix True Condition Rev. Sci. Instrum., vol. 72, no. 2, pp. 1548–1553, 2001.
Left Right Left Right Left Right [7] X. Lin and B.-C. Seet, “A linear wide-range textile pressure sensor
Respondent 1 Normal HighArch HighArch Normal F(-) F(+) integrally embedded in regular fabric,” IEEE Sens. J., vol. 15, no. 10,
Respondent 2 FlatFeet Normal HighArch HighArch F(-) F(-) pp. 5384–5385, 2015.
Respondent 3 Normal FlatFeet Normal Normal T(+) F(+)

140

View publication stats