A Piezoelectric Energy-Harvesting Shoe System for Podiatric Sensing

Rich Meier Student Member - IEEE, Nicholas Kelly Student Member - IEEE,
Omri Almog Student Member - IEEE, and Patrick Chiang Member - IEEE

Abstract— This paper provides an energy-harvesting, shoe- While our solution uses COTS piezoelectric harvesters,
mounted system for medical sensing using piezoelectric trans- there are many other methods for energy harvesting. [6]
ducers for generating power. The electronics are integrated provides a survey of some different methods, including
inside a conventional consumer shoe, measuring the pressure
of the wearer’s foot exerted on the sole at six locations. The electromagnetic, electrostatic, and more. Additionally, [7]
electronics are completely powered by the harvested energy presents a shoe-based energy-harvesting method using di-
from walking or running, generating 10-20 µJ of energy per electric elastomers achieving 1̃20 mJ per step. We chose
step that is then consumed by capturing and storing the piezoelectric harvesters due to their COTS availability, small
force sensor data. The overall shoe system demonstrates that form-factor, and potential for mechanical energy capture.
wearable sensor electronics can be adequately powered through
piezoelectric energy-harvesting. In this paper, we demonstrate a podiatric sensing shoe
system that is powered completely by the movement of
I. I NTRODUCTION the wearer. Off-the-shelf electronics are used for energy-
The continued improvement in size, cost, and power harvesting capability, and to obtain distribution data of foot
of integrated circuits has enabled entirely new classes of pressure. The proposed system is also vertically integrated,
wearable devices. Unfortunately, these new devices are still including not only the hardware, but also the coordinated
limited by current battery technology, which in many cases visualization and database back-end.
is heavy, expensive, and unable to store sufficient power The paper is organized as follows. Section II will highlight
for long-term biomedical sensing. Hence, wearable devices each of the subsystems and their functionality. Section III
powered by an alternative method, such as energy-harvesting, will discuss the preliminary results of this prototype system,
are desirable. and Section IV which will specify some of the further
Of all the locations on a human’s body, the feet experience testing and future capabilities necessary to move this research
the highest levels of mechanical and kinetic energy during forward. Finally Section V will conclude this work.
normal use. Therefore it makes sense to embed energy-
II. S YSTEM D ESCRIPTION
scavenging within a conventional shoe. Furthermore, due to
the large numbers of injuries and conditions associated with A. Energy Harvesting
high-intensity athletic activities, podiatric sensing is a natural The energy-harvesting capability of this system was de-
application for this harvested energy sensor system. signed to maximize energy capture by harnessing multiple
Previously, there has been a limited amount of piezoelec- excitation sources. Since piezoelectric transducers produce
tric, energy-harvested, shoe-mounted bio-sensing systems. electrical energy only by physical deflection, we sought to
However, there has been research on energy-harvesting using harness energy from both foot strikes and bending. As seen in
piezoelectrics, including [1], [2], [3], and [4]. Most notably, Table I and Figure 1, there were two piezoelectric transducers
the Responsive Environments Group at the MIT Media Lab- utilized — a rigid energy-harvester and a flexible energy-
oratory created a system that utilized the energy harvested harvester. The rigid transducer was enclosed in a low-profile,
from two piezo transducers to broadcast RFID signals [1]. custom 3D printed enclosure that allowed it to vibrate freely
Their system utilized custom PZT (Lead zirconate titanate) without breaking. The flexible-energy harvester was placed
and PVDF (Polyvinylidene fluoride) transducers to produce strategically at the ball of the foot to maximize foot strike
1.3 mW continuously (at a 0.9 Hz pace). While their system excitations as well as bending excitations, via downward
generates greater power, our system utilizes limited, com- compression and foot flexion, respectively. Ultimately, the
mercial off-the-shelf (COTS) harvesters to achieve similar goal is to capture otherwise wasted energy generated by the
goals. natural movements of walking, running, and general athletic
There has also been some previous work on mobile, activity.
podiatric sensing (such as gait analysis). Stacy Morris of
MIT developed a wireless, shoe-based, gait analysis system B. Power System
[5]. The system utilized Force sensitive resistors (FSRs) This subsystem contains the power conditioning circuitry
and PVDF transducers for sensing both static and dynamic that allows the shoe system to operate. Since the exact
forces. In our case, we utilize only FSRs to record gradual amount of energy harvested by the piezoelectric transducers
changes in pressure, since our sampling frequency is low. was unknown during the design phase, the power circuitry
Similarly to their implementation, we use a heel-strike sensor was created with operational flexibility in mind. Three dis-
(PZT) to determine when to start sampling the FSRs. tinct operating modes were chosen in order to allow for
978-1-4244-7929-0/14/$26.00 ©2014 IEEE 622
 TABLE I
PRIMARY SYSTEM COMPONENTS

COMPONENT PART NO. DESCRIPTION ACTIVE POWER

Energy Harvesting
Mide Volture - PZT Piezoelectric Element V25W Rigid Vibrational Transducer 20 − 40 µW †
Physik Instrumente Durract - Processed PZT P-876.A11 Flexible Piezoelectric Transducer 5 − 10 µW †
Power Circuitry
Linear Technologies - Integrated Circuit LTC-3588-1 AC-DC - Piezo Power Conditioning 1.75 − 45 µW
Cymbet Enerchip - Integrated Circuit CBC-3150 Solid State Battery & Power Control 11.55 µW
CDE Acrylic Capacitors FCA1210C105M-G2 Low Leakage Energy Storage -NA-
Microcontroller and Communication
Texas Instruments - Microcontroller MSP430FR5739 CPU & Data Storage 6.44 mW
FTDI - Integrated Circuit FT232RQ UART to USB 2.0 49.5 mW
Sensors
Tekscan - Resistive Sensors A201 Flexible Force Sensors 0.098 − 0.99 mW
CUI Inc. - Piezoelectric Diaphragm CEB-35D26 Passive Piezoelectric Sensor -NA-
† Highly dependent on: frequency of steps, user weight, transducer loading, cantilever tip mass, general mechanical stress/deflection.

store the captured piezoelectric energy during Mode-2 until

sufficient charge is accumulated, and then consumed during
Mode-1. A CBC-3150 (Enerchip) provides logic signals as
Pressure Sensors well as a small 50 µAh battery which is used in Mode-
2 (sleep state) to keep the time. During Mode-3 (USB
operation), the power provided by the USB is utilized to
run the system during data download as well as recharge the
solid state battery (Enerchip).

C. Microcontroller and Communication

Cantilever Enclosure The microcontroller subsystem is the main hub for con-
System Enclosure
trolling communication, storage, and processing within the
Bending Harvester
system. The main control signals of the microcontroller
include: sensor ready signals that indicate if the wearers foot
Foot is on the ground and thus whether sensor readings should
Strike Sensor
be captured, power ready signals that indicate when there
Vibration Harvester
is enough energy accumulated to perform data read and
System PCB Velcro
store, and finally USB communication controls that send and
Fig. 1. Expanded view of the shoe system and all integrated components. receive data while in Mode-3.
During the design phase, it was deemed extremely impor-
flexible duty cycling of data capture. Mode-1 characterizes tant that the microcontroller consume ultra-low power, as the
the system in a fully awake state that is capturing sensor energy-harvesting system only produces power on the level
data, and is powered completely from the stored piezoelectric of microwatts. Furthermore, the microcontroller also needed
energy (while going into an additional sleep state between to incorporate at least six ADC channels to allow for parallel
samples). Mode-2 characterizes the system in a sleep state, conversion of the sensor data rather than serial, in order to re-
where the harvested piezoelectric energy is not sufficient to duce the amount of time spent in energy-consuming Mode-1.
capture the sensor data. Both Modes 1 and 2 occur while the A UART interface was also required in order to communicate
user is wearing the shoe system and is exhibiting some kind with the USB chip in our system. Lastly, the microcontroller
of foot movement. Finally, Mode-3 is defined as the non-user was required to incorporate non-volatile data storage, as
mode, where the shoe system is connected via USB for data off-chip data storage (such as Flash RAM) demanded too
download and is no longer capturing sensor data. much power. The specific Texas Instruments microcontroller
In order to power the system in all three modes, the that was used can be seen in Table I. This microcontroller
power circuit uses the two integrated circuits and low leakage provides the ability to store 16 KB of data in on-chip FRAM,
capacitors annotated in Table I. The first chip is a Linear consume only 1.2 µA (3.3 V) when idle, while incorporating
Technologies IC that provides a combination of both an 33 general purpose I/O pins, a UART interface, and 14 ADC
AC/DC rectifier and buck/boost converter. The buck/boost channels. Therefore, it sufficiently meets all of the above
converter is chained to low-leakage acrylic capacitors that requirements. Since FRAM is still an emerging technology,
623
limited sizes are available (maximum of 64 KB); however,
with our data storing implementation (4 B for timestamp, 2
B for each sample) allows for approximately 1000 steps to
be stored (assuming 5 samples per measurement).
D. Sensors
The sensor subsystem consists of the final data-capturing
circuitry, and has two main purposes. The primary function
of the sensor block is to provide an analog signal for each
of the resistive force sensors placed within the insole of
the shoe. Force exertion is converted to a voltage signal
via these sensors, with this data routed to the I/O pins
of the microcontroller for processing and storage. In the
final prototype, we use flexible sensors (shown in Table I)
that allow for measurement of both static and dynamic
forces. Flexibility was important in order to maximize safety,
minimize walking impediment, and prevent the sensors from Fig. 2. Custom GUI visualization of force distribution of the foot while
breaking when in use. walking. Note the location of the 6 sensors.
The second function of this block is to provide an interrupt
signal to the microcontroller, asserting when the foot is
placed on the ground. The sensor chosen to provide this the piezoelectic elements utilized are characterized according
“sensor-ready” signal is a passive piezoelectric sensor. This to their frequency of oscillation, it is difficult to exactly
sensor is ideal because it requires no power or amplification, quantify the energy capture, as human movement consists
and it senses dynamic force only. This latter characteristic of a superposition of many frequencies. Sudden changes
makes it ideal for sensing foot strike and liftoff. in movement can cause both constructive and destructive
interference during oscillation. Despite high variability in
E. Computer Software operating amplitude and frequency, we consistently observed
In Mode-3, all data from the device is transferred to an average of 10 − 20 µJ of energy capture per step. Since
custom PC software via USB. The software is responsible this value is based on counting the capacitor charge, these
for processing the sensor data, storing it in a database, and measurements also include all efficiency and parasitic losses
displaying various data visualizations (pressure map, graph, in the integrated circuits, shown in Table I.
datatable) to the user. Post-processing is performed to convert One example of a charging test is depicted in the graph
the 12-bit ADC values to the corresponding force value in seen in Fig. 3. This shows the results of typical walking
Newtons. Each set of samples is saved with a timestamp movement. The raw piezoelectric voltage waveform (labeled
for later visualization. During analysis, the pressure map and in the color teal) is characterized by an AC oscillation
is the most useful representation (Fig.2), which shows a between −1 and 6 V. The high-frequency vibrational trans-
RGB gradient of the pressure mapped onto the sole. The ducer is seen superimposed on the low-frequency bending
data can be redisplayed in the time-domain using a timeline, transducer. After rectification, the energy is stored on low-
showing the pressure transitions of the foot within the shoe. loss input capacitors that are used to regulate the input to
In addition, when the shoe is connected to the computer the Linear Technologies energy-harvesting IC. As charge
the user can change any programmable settings (such as accumulates, the input capacitors move energy across the
sampling period and sample size) and view/record sensor boost converter to the output capacitors, which are then
data in real-time. regulated by the Enerchip to supply a consistent 3.3 V
operating point for short durations of time. The Enerchips
III. R ESULTS power ready signal (labeled and seen in green) alerts the
A. Power Generation and Delivery microcontroller when enough charge has accumulated on the
The complete prototype, including the energy harvesting output capacitor bank. Hence, the Mode-1 operating state
and power circuitry, was able to sufficiently power the shoe (described in Section II-B) can then be entered to capture
system at different sensor capture rates, depending on the the force sensor data.
activity of the wearer. For typical walking situations, approx-
B. Recorded Measurements
imately 10−20 steps were needed during Mode-2, in order to
have enough energy to enter into Mode-1 and capture sensor The final prototype consisted of six flexible force sensors
readings. During running, duty cycle relating to the number placed at the Calcaneus (heel), Cuboid (lower outside), Nav-
of steps required for each sensor reading was varied between icular (lower inside), the head of the first and fifth metatarsals
1 and 5 steps. These step numbers would likely vary slightly (upper inside and outside), and the head of the proximal
depending on the variance of individual user. In our lab tests, phalanx of the big toe (Fig. 2). Various tests were performed,
a 90.7 kg male of height 1.8 m completed the tests. Because such as jumping, walking, and jogging tests. Jumping tests
624
 IV. F UTURE W ORK

Heel Strike
Waiting for heel
Input Capacitor Bank
Piezo Input While an initial prototype for podiatric sensing using
Output Capacitor Bank energy harvesting has been created, there is still much
Power Ready
6V validation and calibration that needs to be performed in order
to obtain precise (absolute) medical analysis. The sensors
need to be calibrated within the shoe and validated against
conventional methods for podiatric/gait analysis. Similarly,
Signal Voltage (V)

4V the software could be improved to include features requested

by the medical community who may be using the device
and software (e.g., physical therapists or sports scientists).
Other improvements include increasing the energy capture
2V efficiency as well as optimizing the piezoelectric transducers
to capture the most energy possible while minimizing para-
sitic losses in the circuitry. Improving the energy-harvesting
efficiency would enable wireless data collection, such as
0V
Bluetooth 4.0 Low-Energy, thereby enabling smart-phone ap-
1 2 3 4 5 6 7 8 9
plications or wireless, real-time monitoring of foot pressure.
Time (s) V. C ONCLUSION
Fig. 3. A time-domain graph of energy capture in lab walking test. Note, The system described in this paper combines novel energy-
the device is usually in Mode-2, only transitioning to Mode-1 for a brief harvesting techniques with force-based sensors to deliver an
amount of time after heel strike. Piezo energy is consumed in Mode-1 only.
innovative solution to conventional in-lab equipment. The
system is designed to be robust, mobile, and fully embedded
in the patients normal routine, allowing for podiatric analysis
in a variety of environments. Due to the low-volume and low-
maintenance features, the device can be targeted for athletes,
physical therapy patients, amputees, and those with muscular
or nervous system disorders.
ACKNOWLEDGMENTS
We would like to thank Joe Crop and Jacob Postman
in Prof. Chiang’s research group, and Don Heer for their
knowledge and guidance. Thanks to David McKinley for his
advise and hardware support. Finally, we would like to thank
TekScan Inc. for supplying sensors for this research.
Fig. 4. Completed and functional piezoelectric energy-harvesting shoe
prototype (same size and height as unaltered shoe). R EFERENCES
[1] Shenck, N.S.; Paradiso, J.A., “Energy scavenging with shoe-mounted
piezoelectrics,” Micro, IEEE, vol.21, no.3, pp.30,42, May/Jun 2001.
consisted of the wearer jumping up and down for 5 seconds [2] Ishida, K.; Tsung-Ching Huang; Honda, K.; Shinozuka, Y.; Fuketa,
in order to verify accurate data capture. The more useful tests H.; Yokota, T.; Zschieschang, U.; Klauk, H.; Tortissier, G.; Sekitani,
consisted of walking or running for a discrete time interval. T.; Takamiya, M.; Toshiyoshi, H.; Someya, T.; Sakurai, T., “Insole
pedometer with piezoelectric energy harvester and 2V organic digital
These tests were conducted in an urban environment, such as and analog circuits,” Solid-State Circuits Conference Digest of Technical
on pavement or linoleum. The tests spanned anywhere from Papers (ISSCC), 2012 IEEE International, vol., no., pp.308,310, 19-23
10 seconds, to 15 minutes, and were performed in a casual, Feb. 2012
[3] Rocha, J. G.; Goncalves, L.M.; Rocha, P.F.; Silva, M. P.; Lanceros-
conventional daily routine environment. Our recorded results Mendez, S., “Energy Harvesting From Piezoelectric Materials Fully
were as expected in terms of relative force. As the wearer Integrated in Footwear,” Industrial Electronics, IEEE Transactions on,
walked around, the force transitioned from the heel to the vol.57, no.3, pp.813,819, March 2010.
[4] Fourie, D., “Shoe mounted PVDF piezoelectric transducer for energy
toes, with most of the force occurring on the initial impact of harvesting”, MORJ rep, vol.19, pp.66-70, Spring 2010.
the heel. This is in contrast to running or walking down stairs, [5] Bamberg, S.J.M.; Benbasat, A.Y.; Scarborough, D.M.; Krebs, D.E.;
where the force is mostly centered around the ball of the foot, Paradiso, J.A., “Gait Analysis Using a Shoe-Integrated Wireless Sensor
System,” Information Technology in Biomedicine, IEEE Transactions
with less impact on the heel. While our system measures on, vol.12, no.4, pp.413,423, July 2008.
force in Newtons, actual force measurements inside the shoe [6] Paradiso, J.A.; Starner, T., “Energy scavenging for mobile and wireless
were not verified by another device. Therefore, the sensor electronics,” Pervasive Computing, IEEE, vol.4, no.1, pp.18,27, Jan.-
March 2005.
data recorded is currently qualitative, though with further [7] Goudar, V.; Ren, Z.; Brochu, P.; Potkonjak, M.; Pei, Q., “Optimizing
calibration and refinement of our sensor voltage versus force the Output of a Human-Powered Energy Harvesting System With
correlation (currently an exponential best-fit), it is possible Miniaturization and Integrated Control,” Sensors Journal, IEEE, vol.14,
no.7, pp.2084,2091, July 2014.
to obtain accurate, quantitative force data.
625