Vous êtes sur la page 1sur 7

Development of and Applications For a Low Cost,

Open-source Pressure Mat Within a Smart Home


Environment

Daniel H. Goodman
Department of Media Arts and Sciences
MIT Media Lab, Changing Places Group
Cambridge, MA, USA
dhgood@mit.edu


Abstract The Changing Places group of the MIT Media Lab
recently completed a prototype for a small-scale (200ft
2
)
transformable apartment to address some of the problems related to
high-density urban living. The small size of the space paired with the
density of technology creates an ideal environment for continuous
sensing of the user. Of specific interest to the author is the recording
of gait data to predict likelihood of fall as well as document
precursors to and progression of joint diseases such as osteoarthritis.
To allow for easy integration into the technology ecosystem of the
CityHome, the author designed and constructed a 22x32 pressure
mat with 1.25x1.25 resolution making use of off-the-shelf hobbyist
components and the Arduino microcontroller system. Pairing the
mat with the backend software of the CityHome, the system enables
real time display of mat pressure readings as well as speculative
simulated windows related to joint health information addressable by
voice command.
KeywordsGait analysis, smart home, MIT Media Lab, pressure
sensor, falls, osteoarthritis
I. INTRODUCTION
The CityHome prototype constructed by the
Changing Places Group will be used as a testing laboratory for
hardware and software applications intended for deployment
in a smart home. Though the possibilities for applications
can fall under many different categories, two very important
areas of research are related to rehabilitation and geriatrics.
In 2013, the CDC estimated that one in three adults
over the age of 65 falls [2]. Many families have first hand
experience with grandparents or elderly parents falling and
severely injuring themselves. Injuries range in severity from
bruises to hip fracture and head trauma. The cost to treat all of
these injuries exceeded $30 billion in 2010

[2]. Many of these
people are ambulatory before the fall, but the injury and
subsequent healing often lead to long term negative impacts
on quality of life and even risk of death. For those who do
recover, many lose confidence in their stability while walking,
and as a result avoid ambulation.
In addition to falls, joint disease, and in particular
osteoarthritis, is becoming more prevalent in the obese and
elderly population. Osteoarthritis affects about 27 million
people in the US and causes pain and limited mobility in the
affected joints [3, 5]. It develops gradually and typically
progresses over many years. There is currently no cure;
treatments include physical therapy, weight loss, or in extreme
cases joint replacement surgery.
For both falls and joint disease, several warning signs
have been identified that are manifested as alterations in the
normal gait. Factors such as reduced gait speed, increased
stride time, and increased stride length have been linked to a
higher likelihood of falling while lower walking speed, lower
cadence, shorter step length, and an adjusted gait have been
linked to gonarthritis and osteoarthritis [1, 4, 6, 7]. A pressure
and location sensitive carpet, especially in the home
continuously monitoring the occupant, could continuously and
unobtrusively capture a rich set of gait data. The data can be
analyzed in real time and warning markers could be identified.
Continuous monitoring of walking could be a boon
for doctors and family members, so that any early warning
signs could be identified and treatments could begin earlier.
The home is an optimal location for a gait-sensing device
embedded in the floor/carpet, because recordings can be taken
daily under similar conditions. The CityHomes computer
system, for example, could run continuous analyses on the
incoming data and alert appropriate parties should any
warning signs appear.
To collect spatial and force data from walking, a
matrix pressure mat would need to be employed. There are
several companies that design pressure sensing mats for
orthopedic applications, and though the resolution is more
than adequate for gait analysis, the systems are often closed-
source, requiring proprietary hardware and software to use
[13, 14]. Large pressure mats, on the order of a floor carpet,
are also prohibitively expensive and preclude use outside of
medical facilities and university labs. In an effort to create a
practical and low cost sensing platform for the home, the
author set out to develop a pressure mat that could easily
integrate with the CityHomes operating system.
Once the pressure mat was designed, built, and
running through the CityHomes computer, the author created
speculative interfaces, which served as a demonstration of the
value that continuous gait monitoring could bring to the user.
Questions such as what is my likelihood of falling? or how
did I walk this week? could be posed to the system, and with
proper analytics of the incoming data, an answer could be
generated.
II. METHODS
A. The CityHome

Fig. 1. The CityHome Prototype

Note the central furniature module. This module contains the bed, desk, and
computer system for the home. The RGB lighting and projection system can
be seen above the living room.
The CityHome is a scalable strategy for creating highly-
efficient technology-enabled apartments that can help make
living more affordable, productive, enjoyable and creative for
urban dwellers [18, 19]. By combining transformable
mechanical elements, a centralized operating system and
compatable peripheral devices, the CityHome aims to optimize
and enhance the experience of living in a small urban space.
Transformable furniture allows the user to reconfigure the
room on demand and to fit the specific requirements at that
point. The bed and desk fit into a main module and are
deployed when needed. The module itself can move about the
apartment space, creating a larger bathroom, or a large living
room. Using the devices and software of the home, developers
can design hardware and software apps to enrich the
customizibility and interaction for the user. A CityHome app
store can serve as a depository of apps for the home, to be
downloaded and used with minimal effort on the part of the
user. Hardware add-ons to the native hardware of the home
can pair with specific apps or serve as additional building
blocks for the home device ecosystem. This is where the
pressure sensor-enabled carpet would fit.
B. The Pressure Mat
The pressure mat operates under the same principles as a
Force Sensitive Resistor or FSR. A FSR typically consists of
two strips of a conductive material sandwiching or contacting
a layer of material with a force-dependent resistance, see Fig.
2 [10, 12, 17]. In this case, the material was Velostat, which is
manufactured by 3M [11]. Velostat or similar materials
consist of a polymer impregnated with carbon. As pressure is
applied to the FSR material, the carbon particles become more
densely packed, resulting in lower resistance, which can be
easily measured as a voltage change.

Fig. 2. Force Sensitive Resistor and its Construction



The Arduino has the ability to write and read analog
and digital signals. For the case of reading an FSR, the
Arduino applies 5V (digital write) to one terminal, and reads
the voltage at the other terminal (analog read) [8, 9].
Applying a force to the pad will vary the resistance and
therefore the recorded voltage. Generally, a pull down
resistor is used to divide the output voltage and bring the
readings into the measuring range of the Arduino before being
collected.

The schematic for a typical FSR seen in Fig. 2 is
shown below. Note the polymer is denoted as Velostat in the
diagram.

Fig. 3. Schematic For a Single FSR Connected to an Arduino

5V comes from the Arduino digital pin and A0 is the analog pin. R represents
the pull-down resistor, usually 10k! for a single pad.

In order to collect data regarding the location of the
force, a different configuration is needed. There are still two
conductors sandwiching the polymer, but this time there are
rows and columns of conductive material on either side of the
sheet. Each row is connected to a digital write pin on the
Arduino, and each column is connected to an analog read pin.
By writing one row high (5V) and reading each of the
columns, a scanning read of the entire mat can be achieved.
This method gives row*column unique values [17].

Tekscan, a well-known pressure sensor manufacturer,
uses this collection method in their pressure sensors. A key
difference is that they use an inverting amplifier before
reading the signal [14].

Fig. 4. Schematic For Row and Column Configuration of the Pressure Mat



Fig. 5. Schematic from Tekscan For Reading a Pressure Mat

Tekscan utilizes an inverting amplifier before reading the output signal. For
the prototype described, the author uses pull down resistors [14].

Two iterations of the pressure mat were constructed,
one that was ~12x12 and another that was 20x32. An
Arduino Uno controlled the smaller version, and an Arduino
Mega controlled the larger version. For the smaller iteration, a
0.25 thick square of acrylic served as the base. Strips of
copper foil tape were laid down so that there were five
columns. Using a 1/16 square of acrylic as the stepping
surface, the five copper rows were placed underneath. A
square of polymer was placed in between the intersecting
conductors, and the termini of the strips were connected to
their respective pins on the Arduino. For this configuration,
the data read by the Arduino consisted of a string of 25 values
between 0 and 1023 (analog read range) separated by a space.
The resolution was fairly poor, with one node every ~2.5.

Fig. 6. The First Pressure Mat Prototype

The black material is the Velostat. The rows and columns can be seen above
and below the Velostat.

The second iteration was closer in area to a small rug
and was sized so that at least two steps could be captured.
Instead of acrylic, the two main surfaces were thick paper.
One of the lessons from the earlier design was that due to the
more rigid nature of acrylic, any applied force was distributed
over a larger area, even if the force wasnt being applied at the
surrounding points. Paper is more flexible, so the applied
forces dont bleed into neighboring nodes as much. The new
pad had fifteen columns by twenty-three rows. The number of
analog read pins limited the number of columns. Since the
Arduino Mega has forty-eight digital pins, there were more
than enough pins available for the rows [8]. The string of data
was now 345 values long. The resolution of the new mat was
about 1.25x1.25.















Fig. 6. Two Images of the Second Mat Prototype



C. Visualizing the Data with Processing
Processing is a visual programming language, with a very
similar syntax to Arduino (in fact, the Arduino syntax was
modeled off of Processing). Data can be read by the program
from the Arduino and plotted, manipulated, and visualized etc.
[20].

The scanning read method the Arduino uses to collect the
mat data produces an array of values that is row*column long.
Processing arranges the data from the array into a matrix
representing the spatial position of the specific value. The
matrix is then plotted where a square represents a node of
column and row intersection. Each square is colored on a
scale of black to white depending on the magnitude of the
recorded force on that node.










Fig. 7. Raw Visualization of Large Pressure Mat, Unloaded and Loaded

The first image is the unloaded case. In the second image, the author places
his fingers on the mat; each of the black squares indicates a finger contact
point.

The above method works well for visualizing point
forces, but a way to construct footsteps from the squares was
needed. Using the OpenCV functions within the Processing
environment, the output in Fig. 7 is analyzed, and clusters of
high force (indicating a step or distributed force) are pulled
from the surrounding data. Using another function, the
centroid of that specific cluster can be plotted and its
coordinates saved. The centroid of the cluster approximates
the center of the footstep. By logging the coordinates of each
centroid, the path of a user across the mat can be plotted. In
addition to x and y coordinates of the step, the average forces
present as well as the time between steps can also be recorded.
These data can be used to analyze the gait characteristics of
the user.

Fig. 8. OpenCV Isolated Clusters of Force and their Centroids

The centroids are plotted once the user reaches the end of the mat, creating a
walking step pattern. In the second image, the author used his hand to apply
force and create the step path.
D. Interfacing with the Gait Data
This section is currently speculative, as there is insufficient
accuracy in the pressure mat as well as insufficient training
data to generate prognoses related to gait. This section
assumes an accurate, floor carpet-sized pressure mat recording
and analyzing gait data in real time and for several months.
Given these assumptions, the author generated simulated
outputs corresponding to potential user queries. Ideally, if
ones home is constantly collecting walking data, he/she
would want to be able to access it, and be shown trends and
warnings if necessary. Using the CityHome, the user is able
to ask for specific results, which the house could then project
against the wall for the user to see. In the CityHome
prototype, the author linked these simulated result pages to
specific user vocal commands.

Figure. 8. CityHome, how was my walking this week?


By recording the gait data for the week, the
CityHome can generate plots showing the trends in various
gait features. The user can query the system using natural
speech to obtain an analysis.

Fig. 9. CityHome, how am I walking?


This query is particularly interesting because it is
qualitative. An answer to this question requires a large dataset
(in this simulated example, it was a year) and advanced
analytics that can identify trends in gait and link them to
clinically validated warning markers. In this example, the
CityHome notices the user has some signs of emerging
osteoarthritis. It suggests ways of preventing its progression,
and reassures the user that the physician has been sent the data
as well. [16]. The user now can take an active role in
checking in on gait health and be notified of problems as well
as ways to improve.

Fig. 10. CityHome, what is my likelihood of developing osteoarthritis?


This question is exciting to ponder because with
sufficient data and analytics, your home could accurately
predict your chances of developing joint disease based on your
walking pattern. In essence, your gait becomes a trackable
biomarker, just like weight, blood pressure and other
physiological parameters. In this theoretical situation, the
CityHome would present a screen showing a percent and a
message. The message could give encouragement and
reinforce good behavior. Eating well, staying active etc. can
be promoted as ways of reducing the likelihood of joint
disease.

Fig. 11. CityHome, what is my likelihood of falling?


Similar to the previous question and Fig. 10, an
answer to this question would involve a likelihood paired with
a message. One of the problems associated with falling is that
often the person loses confidence in walking. Obtaining
feedback and encouragement from this system could have a
positive impact on the user continuing to ambulate.
Though much work is needed in the development of
living space-scale sensing, generating predictive algorithms,
and analyzing the gait data for statistically significant trends in
real-time, the idea of the home providing diagnostics and
coaching is very exciting and worth developing further.

III. RESULTS AND DISCUSSION
A. The Constructed Pressure Mat
The 22x32 mat proved to be directionally successful in
the trajectory to develop a low cost system. The overall cost
was less than $100 compared to the several thousand dollar
cost for a similarly sized professional mat. Refining the design
can bring the price down further while improving the
resolution and repeatability. For example, adding a multiplexer
could yield 0.5x0.5 or 0.25x0.25 resolution with the same
Arduino chip.
The current design with the 1.25x1.25 resolution is able
to recognize 345 unique sites of force. The collected data can
be sent to any number of programs via serial for analysis and
visualization. Keeping the interface protocol as standard serial
will allow more people to interface with the data and develop
useful apps.
Unfortunately, while the mat was able to recognize point
forces very well, steps were harder to consistently identify. To
identify these steps using the OpenCV functions, a large
amount of filtering was needed to deal with the noisy signal.
This prevented statistically significant results to be obtained
from from the current mat. Another challenge was that the mat
had difficulty recording clusters when a user stood on the mat.
The electronics and circuitry need more development to
improve the ability to capture higher distributed forces.
B. Implications of Continuous Gait or Biometric Sensing
within a Smart Home
Looking at the CityHome as a model for potential future
technology-driven homes, the author believes there is a
significant opportunity to collect and analyze user data in such
a space. Not only can the data be used to optimize the users
preferences and home environment, but it could also be utilized
in a number of related applications.
In the scope of healthcare, these data can be used for
patient monitoring, diagnostics and providing reassurance.
Having a floor-sized pressure sensing carpet could provide
value to both patients and healthcare providers. With robust
algorithms and sufficient resolution, gait data collected
continuously could be used to identify likelihood of falling or
likelihood of developing debilitating conditions such as joint
disease. These innovations could help millions of people and
save billions of dollars by reducing the number of preventable
falls, and identifying diseases such as osteoarthritis early
enough to treat and slow further progression. More people can
remain mobile longer.
In addition to preventing falls or joint disease, continuous
gait monitoring could provide encouragement to those who
have already fallen or are suffereing from musculoskeletal
disease. The ability to track progress and receive
encouragement from the home could have a positive
psychological impact on the user. Returning confidence in
ones walking, and providing suggestions for improving
symptoms could encourage increasing mobility and taking a
more active role in their health management.
The continuous monitoring doesnt have to be limited to
gait. Imagine how heart rate monitoring or non-invasive
cholesterol measurement, or caloric intake tracking could be
used. Any ailment that requires a large dataset to track its
progression could be monitored using embedded sensors
within the home. Doctors checkups could be transformed
because more data could be collected around the house and
analyzed faster than in any doctors office visit. The role of the
primary care doctor may shift from collector of data to
manager of data thanks to the CityHome system.
In the not too distant future, it is not unreasonable to
imagine a scenario where physiological sensors, algorithms
and analytics could be used to predict disease better than a
doctor.
The author is excited to continue exploring continuous
biometric monitoring in the home. With large datasets of
biometric data, the diagnosis and treatment of many chronic
illnesses could be enhanced tremendously. Diseases can be
identified earlier, treatments could begin earlier, and quality of
life for millions can be improved.
C. Future Work
Several improvements need to be made to the pressure mat
before it can reliably collect gait data. The electronics need to
be improved so that higher forces on the order of body weight
can be collected, because the existing design had trouble
identifying clusters of body-weight force. Additionally, the
material choices in the pressure mat need further investigation.
Velostat and similar materials have large hysteresis and
relatively slow settling times. For more rapid and accurate
force measurements, other materials may need to be used.
The author wishes to eventually design modular pressure
tiles that can be connected together to create larger mats. This
way, an arbitrary-shaped floor can be covered in pressure
sensing carpet. The Arduino can be reduced to a very small
PCB and with the aid of multiplexors, cut the number of output
pins needed to control a single mat.

IV. CONCLUSION
In this paper, the author discusses a design for an
open-source, low cost pressure mat carpet system and visions
for continuous gait monitoring within a smart home
environment. The constructed pressure mat was a fraction of
the cost of similarly sized professional mats and had
reasonable resolution. It was able to identify 345 unique
pixels and the data was easily collected and visualized with
Arduino and Processing. Unfortunately, the accuracy of the
mat was not good enough to reliably and consistently identify
steps. Further development of the system is needed before
running feasibility and other user studies and data collection
experiments. Though simulated, the outputs for the user
questions regarding walking and joint disease predict a very
exciting world where the home will be able to identify
diseases and suggest ways to
improve the situation. By continuously analyzing biometric
data, a smart home environment such as the CityHome could
predict joint disease, fall likelihood or any number of other
conditions earlier on. Alerting the user, users family, or
doctor could prevent disease progression and improve quality
of life.
ACKNOWLEDGMENT
The author would like to thank Professor Hugh Herr for the
suggestion to explore this area of monitoring within the home.
The author would also like to thank Jean-Franois Duval and
Carson Smuts for their assistance in debugging and testing the
prototype pressure mat.
REFERENCES

[1] Brach JS, berlin JE, et. al. (July 26, 2005). Too much or too little step
width variability is associated with a fall history in older persons who
walk at or near normal gait speed. The Journal of Neuroengineering and
Rehabilitation. Retrieved from
http://www.ncbi.nlm.nih.gov/pubmed/16042812.
[2] Falls Among Older Adults: An Overview. Centers for Disease Control
and Prevention. Retrieved from
http://www.cdc.gov/homeandrecreationalsafety/falls/adultfalls.html
[3] Osteoarthritis. Centers for Disease Control and Prevention. Retrieved
from http://www.cdc.gov/arthritis/basics/osteoarthritis.htm.
[4] Montero-Odasso M, Shapira M, et. al. (2004). Gait velocity in senior
people. An easy test for detecting mobility impairment in community
elderly. The Journal of Nutrition, Health & Aging. Retrieved from
http://www.ncbi.nlm.nih.gov/pubmed/15359349.
[5] Osteoarthritis. Arthritis Foundation. Retrieved from
https://www.arthritis.org/conditions-treatments/disease-
center/osteoarthritis/.
[6] Bejek Zoltan, Robert Paroczai et. al. (2006). Gait Parameters of Patients
with Osteoarthritis of the Knee Joint. Physical Education and Sport.
Retrieved from http://facta.junis.ni.ac.rs/pe/pe2006/pe2006-02.pdf.
[7] Sarmini, Mahmoud M.D. Understanding Normal & Pathological Gait.
Retrieved from
http://www.medschool.lsuhsc.edu/physical_medicine/PPT/Normal_Path
ological_Gait.ppt.
[8] Arduino Mega. Arduino. Retrieved from
http://arduino.cc/en/Main/arduinoBoardMega.
[9] analogRead(). Arduino. Retrieved from
http://arduino.cc/en/Reference/analogRead.
[10] Fraden, Jacob. Handbook of Modern Sensors. 3rd ed. N.p.: Springer,
2004. Print.
[11] 3M Velostat Conductive Bags and Films. 3M United Kingdom.
Retrieved
from.http://www3.3m.com/catalog/uk/en002/electronics_mfg/static_con
trol_packaging/node_JV20CN7XFBbe/root_K3BHNB8005gv/vroot_5K
R9K2C2KZge/gvel_JQGPTT9723gl/theme_uk_staticcontrolpackaging_
3_0/command_AbcPageHandler/output_html.
[12] Force Sensitive Resistor (FSR). Adafruit. Retrieved from
https://learn.adafruit.com/force-sensitive-resistor-fsr.
[13] http://www.tactonic.com/
[14] http://www.tekscan.com/tekscan-technology
[15] Abellan van Kan G, Rolland Y, et. al. (2009). Gait speed at usual pace
as a predictor of adverse outcomes in community-dwelling older people
an International Academy on Nutrition and Againg (IANA) Task Force.
The Journal of Nutrition, Health & Againg. Retrieved from
http://www.ncbi.nlm.nih.gov/pubmed/19924348.
[16] What is Osteoarthritis? National Institute of Arthritis and
Musculoskeletal and Skin Diseases. Retrieved from
http://www.niams.nih.gov/Health_Info/Osteoarthritis/osteoarthritis_ff.as
p.
[17] Woven Pressure Sensor Matrix. How to Get What You Want. Retrieved
from http://www.kobakant.at/DIY/?p=4296.
[18] http://www.media.mit.edu/research/groups/changing-places
[19] http://cp.media.mit.edu/research/67-cityhome
[20] http://www.processing.org/reference/

Vous aimerez peut-être aussi