IMTC 2007 — Instrumentation and Measurement
`Technology Conference
`Warsaw, Poland 1-3 May 200
`
`Design of a Low-cost Physiological Parameter Measurement and Monitoring Device
`
`G. Sen Gupta’, S.C. Mukhopadhyay”, B.S. Devlin’ and $. Demidenko”
`'School of Electrical and Electronic Engineering, Singapore Polytechnic, 500 Dover Road, Singapore
`"Institute of Information Sciences and Technology, Massey University, Palmerston North, New Zealand
`Email: s.cmukhopadhyay@massey.ac.nz, SenGupta@sp.edu.sg, b.s.devlin‘@massey.ac.nz, s.demidenko@massey.ac.nz
`
`Abstract — In this paper we present the design of a low-cost system
`that can be used to monitor physiological parameters, such as
`temperature and heart rate, of a human subject. The system consists
`of an electronic device which is worn on the wrist andfinger, by an
`elderly or at-risk person. Using several sensors to measure different
`vital signs, the person is wirelessly monitored within his own home.
`An impact sensor has been used to detectfalls. The device detects if
`a person is medically distressed and sends an alarm to a receiver
`unif that is connected to a computer. This sets off an alarm, allowing
`help to be provided to the patient. The device is battery poweredfor
`use indoors. The device can be easily adapted to monitor athletes
`and infants. The low cost of the device will help to lower the cost of
`home monitoring ofpatients recovering from illness. A prototype of
`the device has been fabricated and extensively tested with very good
`results.
`
`physiological
`—
`Keywords
`transmission, home monitoring
`
`parameters,
`
`SERSONS,
`
`wireless
`
`I.
`
`INTRODUCTION
`
`Manyelderly people dread the idea of being forced to live
`with their adult children, or
`in a rest home or in other
`Sheltered
`living
`arrangement.
`They want
`to
`live
`independently and keep control of their own lives. Yet at the
`same time they know there is a high risk of injury or even
`death because of a fall or stroke. With the population aging
`in most developing countries, there will be more and more
`elderly people living alone in future. Such people need to be
`monitored continuously and provided with immediate
`medical help and attention when required.
`The cost of hospitalization is ever increasing, so 1s the cost
`of rehabilitation after a major illness or surgery. Hospitals are
`looking at sending people back as soon as possible to recoup
`
`at home. During this recovery period several physiological
`parameters need to be
`continuously measured. Hence
`telemedicine and remote monitoring of patients at home are
`gaining added importance and urgency [1-3]. Today,
`the
`progress in science and technology offers miniaturization,
`speed, intelligence, sophistication and new materials at lower
`cost. In this new landscape, micro-technologies, information
`technologies and telecommunications are the key factors in
`inventing devices to assist mankind. Patients are being
`monitored using a network of wireless sensors [4]. A system
`to monitorthe overall health of welfare facility residents, who
`need constant care, has been reported in [5]. This system [5]
`has been designed with wireless sensors, wireless repeaters
`and a host computer. The system consists of a piezoelectric
`
`1-4244-0589-0/07/$20.00 ©2007 IEEE
`
`1
`
`sensor, a 2 axis accelerometer, a microcontroller and a low
`power
`transceiver.
`It
`records
`respiration activity and
`indicators of posture for 24 hours. These data are transmitted
`to the wireless repeater by the transceiver. The wireless
`repeaters, which are installed throughout the welfare facility,
`send data, including the repeater's ID, to the host computer.
`The ID is used to detect the resident's location in the welfare
`facility. The host com puter stores the data, which can be used
`to analyze the resident's overall health condition. When the
`resident is in an emergency situation, such as falling or in an
`inactive state for more that
`the allotted time,
`the host
`computer automatically alerts the situation to the care staff by
`an alarm sound and also by mobile phone. However, all
`reported systems are relatively expensive.
`In the back drop of the
`importance of continuous
`monitoring of vital physiological parameters of a patient, our
`research was undertaken to design a
`low-cost
`smart
`monitoring device. It aims to provide peace-of-mind to users
`who have medical problems, but are not placed mn a hospital
`for monitoring. Caregivers, who look after patients with
`mental or physical disabilities, can use this device when they
`are not able to visually supervise the patients.
`Currently there are monitoring products in the market that
`are aimed to provide emergency assistance to senior citizens,
`rehabilitation
`patients,
`and medically
`or
`physically
`challenged individuals, but these have limitations. St John’s
`and Medic Alert’s Lifelink™ [6] allows the user to set off an
`alarm manually if they are under medical stress, which will
`then dial designated contact phone numbers. The fundamental
`problem with this system is that when medical emergencies
`happen to the user, they are often unconscious and unable to
`press an ‘emergency alert button’. There is ne product on the
`market which does not require manualactivation of the alarm
`and monitors a user’s vital signs smartly. This is the novel
`design goal of the work presented in this paper.
`The reported device consists of a wrist strap and a finger
`glove. This allows the sensors to be mounted around the wrist
`and finger. A battery and a microcontroller, with built-in RF
`transceiver, are mounted within the wrist strap as well. In
`Section II we present the complete system overview. All the
`sensors are explained in Section IIT. The hardware details are
`in Section IV and the algorithms in Section V. The prototype
`and test results are discussed in Section VI. The paper ends
`with a discussion on future developments
`
`Apple Inc.
`APL1045
`U.S. Patent No. 8,923,941
`FITBIT, Ex. 1045
`
`Apple Inc.
`APL1045
`U.S. Patent No. 8,923,941
`
`FITBIT, Ex. 1045
`
`

`

`Il. SYSTEM OVERVIEW
`
`B. Heart Rate Sensor
`
`The system has been designed to take several inputs from
`a human subject to measure physiological parameters such as
`temperature and heart rate. Figure 1 shows the functional
`block diagram of the system hardware. The inputs from the
`sensors are processed andthe results transmitted to a receiver
`unit, which is connected to a computer placed in the home,
`using Radio Frequency (RF) wireless
`technology. The
`receiver unit decodes and analyses the data. If it is inferred
`that the person is medically distressed, an alarm is generated.
`The design is modular which makes it rather easy and straight
`forward to add extra sensors for measuring and monitoring
`other parameters. The hardware blocks are explained in full
`details in a later section.
`
`A custom heart rate sensor was designed to read the
`patient’s beats per minute (bpm). The designed sensor is very
`small and inexpensive. The technique used to measure the
`heart rate is based on near-infrared (NIR) spectroscopy. NIR
`spectroscopy involves using light in the wavelength of 700-
`900nm to measure blood volume. At these wavelengths most
`tissues do not absorb light — other than hemoglobin Cwhich is
`what we are interested in). This allowed for designing a non-
`invasive and low cost method of measuring the pulse. A
`silicon phototransister and a GaAs infrared emitting diode
`were used in the sensor, moulded into a flat side-facing
`package. The amount of light
`that was detected by the
`phototransistor varied with the patients heart pulse, as the
`amount of absorbed IR light changed with the flow of blood,
`which is directly linked to the heart rate. This signal was then
`amplified,
`filtered, and sent
`to the microcontroller to be
`analyzed. The heart rate sensor was mounted in the finger
`glove as this position proved to give the best response.
`
`C. Impact Sensor
`
`
`An ADXL311 accelerometer was used as an impact
`sensor. It provides a 2-axis response, measuring accelerations
`up to +/- 2g.
`It was fitted into the wrist
`strap. The
`accelerometer provides an analog voltage, the amplitude of
`which is directly proportional to acceleration. This signal was
`scaled down to bring it within the acceptable input range of
`the micro-controller, and then analyzed. Software algorithms
`were used to detect sharp impacts, while allowing slower
`movements, such as walking, to be ignored. The purpose of
`this sensor was to detect sudden impacts that could indicate
`the patient had fallen over.
`
`
`
`IV. HARDWARE DESIGN
`
`The hardware was built in three separate blocks. A sensor
`card was designed to house
`the
`temperature
`sensor,
`accelerometer and the connections for the NIR emitter and
`detector. A separate analog card was designed for all the
`analog processing circuitry needed for the sensors, primarily
`for processing the heart rate signal. The temperature and
`acceleration output was fed directly to the micro-controller.
`The micro-controller was mounted on a separate card which
`also had the antenna connection. The cards were connected
`by ribbon cables within the wrist strap. Figure 2 shows the
`circuit schematics of the sensor units and Figure 3 shows the
`analog processing circuit for measuring heart beatrate.
`
`A. Lock-In Amplification for heart rate measurement
`
`the infra-red emitter
`The micro-controller modulates
`signal at 1 KHz through a pn transistor (Figure 2). This is
`then mixed with the signal obtained from the TR sensors
`(back scattered light). This technique, known as Lock-In
`Amplification
`[9],
`involves phase
`sensitive
`detection.
`
`FITBIT, Ex. 1045
`
`
`
`Impact Sensor
`(Accelerometer)
`
`et
`2.4GHz RF
`
`J
`
`
`La Mixed-signal
`
`
`
`Temperature
`ce
`micro-controller
`Micro-controller
`ua
`with RF
`Sensor
`‘with RF ee PC
`Circuitry eo Transmitter
`Receiver
`eee
`AnalogSignals
`
`
`
`
`
`Heart Rate
`Sensor
`Circuitry
`
`: \
`
`Fig. 1. Functional block diagram of the system hardware
`
`Ill. SENSORS AND INTERFAC] 23]
`
`The system consists of three sensors- a temperature sensor,
`heart rate sensor, and an impact sensor. All
`the sensor
`circuitries used in the design generate analog voltages which
`
`are fed to the ADC (Analog-to-Digital) inputs of the micro-
`
`controller. The ADC inputs
`are
`time-multiplexed and
`sampled at different rates. The description of individual
`sensors follows.
`
`A, Temperature Sensor
`
`The temperature measurement is done using a LM35 [7]
`precision integrated-circuit temperature sensor. It provides an
`accuracy of +/- 0.25°C within the desired temperature
`measurement range of 20-40°C. It has a very low current
`drain of 60 WA. This sensor is mounted within the wrist strap,
`positioned in such a way that it 1s in contact with the skin,
`allowing it to measure the external temperature of the skin.
`From the skin temperature, the body temperature is estimated.
`Because an exact measurement of body temperature is not
`required, this method is suitable. Rather, relative changes are
`monitored within set thresholds, which set off the alarm. This
`allows the device to detect changes in body temperature that
`could indicate the patient 1s undergoing any of the following
`conditions:
`trauma,
`injury, heart
`attack,
`stroke, heat
`exhaustion, and burns [8]. The temperature sensor is sampled
`once every 3 seconds.
`
`FITBIT, Ex. 1045
`
`

`

`
`Accelerometer
`
`
`
`[Swe
`\ SWB
`
`
`
`=
`
`=
`
`FlareSlart
`|
`i
`oA
`i
`|
`
`as
`Ceriar
`IR signal traneeliter
`1¥ue
`ari geceiver
`Square
`
`:
`5
`ane
`Hn
`=
`Yi
`YeuoxLaa
`zy
`Weg. OM
`|
`
`
`
`aRlUs
`
`Tt
`
`
`“2301
`
`
`Photos Sex
`
`
`
`
`Paes
`ug OMEN
`
`ARS
`SSaR
`
`CT
`
`Connection to saalog poccessing
`_
`~ Toth”
`
`Fig. 2. Sensor Units
`
`
`Analog switch demodulating}
`
`
`8Cea
`
`aL
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`TLD
`
`ot
`weed Bp
`wap
`ey
`
`mth
`
`ie
`
`2c
`
`ADEwitch,
`
`
`
`
`
` 23M
`
`
`oe
`
`
`
`ned
`
`
`ics
`
`C3
`
`Low-Puge filer
`
`AGcouple and amplification (12.6 gain}
`
`Sigurt Hawerter {x1 Gairy
`
`Fig. 3. Signal conditioning for Heart Rate Monitoring
`
`The signal of interest (1-2 Hz) is moved into a much higher
`BeckGealier? Last
`i Te
`f2he)
`frequency band together with the noise that is picked up by
`ADC
`
`
`oes, tyechemus|f PERE
`the sensors. Demodulation is performed by an analog switch,
`EER ore
`ampWiade
`#
`Low pass her
`3
`
`enmluiator
`§
`which demodulates at the same frequency and in phase with
`the 1 KHz modulatmg signal. The resulting signal is then
`passed through a low-pass filter with a cut off frequency of
`6Hz. This allows the signal of interest to pass through, while
`filtermg out all noise that origmated from the sensor. This 1s
`illustrated in figure 4.
`
`| Amplification |
`
`Fig. 4. Lock-In Amplification for heart rate measurement
`
`B. Controller Card
`
`The controller used in the wrist strap unit (and also in the
`receiver unit) is a Nordic nREF24E1. This takes inputs from
`
`the sensor circuits in the form of analog voltages. Each sensor
`has a dedicated ADC channel which is multiplexed by the
`microcontroller. Each sensor’s signal
`is sampled at a pre-
`defined rate,
`through interrupt-driven algorithms. This
`microcontroller was chosen because of its small footprint
`(6mm x 6mm),
`low power
`consumption,
`and built-in
`
`FITBIT, Ex. 1045
`
`FITBIT, Ex. 1045
`
`

`

`transceiver. The microcontroller is responsible for producing
`the carrier signal that modulates the IR ernitting diode, which
`is used in demodulation too. It is powered by a 9V battery,
`which ts regulated down to 5V. The controller card contains
`the microcontroller, crystal oscillator, RF antenna connection,
`and header pins which connect analog and digital ports to the
`other components of the unit
`
`CO. Receiver Unit
`
`The hardware of the receiver unit 1s housed in a plastic
`
`box and consists of a nREF24E1 microcontroller, antenna, a
`5¥ AC adaptor and serial interface port. The receiving unit is
`connected, via a RS232 serial port, to a personal computer
`(PC) and is constantly receiving information about
`the
`patient's medical status A program, running on the PC,
`receives the packetized information from the serial port,
`decodes the packet and then displays this information on the
`PO monitor. The program offers several options to set
`thresholds of the allowable range of heart
`rate,
`skin
`temperature and impact sensor. When the readings from the
`patient rove outside of the set range, an alarrn is raised. The
`software can depict
`the change in data over
`time in a
`graphical plot. One receiver unit can be interfaced with
`multiple wrist units. In the software, one can tab through the
`information received from each wrist unit.
`
`2D. Comumurication
`
`Communication bebyeen the wrist units and the receiver
`
`unit is wireless, powered by the nREF24E1, and transmitted
`in the unlicensed 24GHz frequency band. Information is
`gathered every 3 seconds from the sensors and then encoded
`into a packet. Each packet
`is 6 bytes long, composed as
`shown in figure 5.
`
`Start of File
`
`a—Bit balanced (10101070)
`
`6 Bytes Long
`
`End of File Bit balanced (10101010)
`
`Skin Tamperature
`
`Impact Information
`
`
`is sent by the nREF24E1 using a
`The data packet
`technology termed ShockBurst™. This allows data to be
`clocked into a FIFO buffer at a low data rate, and then
`transmitted all at once at a very high rate. This lowers power
`consumption by minimizing the amount of receiving and
`
`transmittmg time Data is transmitted by this method at
`IMbps. This also reduces the risk of a data collision with
`other wrist units, or devices operating m the sare band.
`
`Vo SOFTWARE AND ALGORITHMS
`
`the
`on
`implernented
`were
`algorithms
`several
`from the
`microcontrolla. These algcritms take inputs
`sensors and process them into meaningful information.
`
`A, Heart Rate Algorithm
`
`The heart rate signal is an oscillating signal of 1 to 2Hz. This
`
`is sampled by the microcontroller at S0Hz, and then averaged
`over 5 samples, which gives the algorithm an effective data
`input rate of 10 sarnples per second. This 1s stored in a buffer
`which is 32 bytes long. Once this buffer 1s full, the algorithm
`scans though each value in the buffer to compute the period
`of the signal (see figure 6). Smncethe signal is not smusoidal,
`rather has distinct peaks corresponding to the contractions in
`the heart, the period is measured by marking transitions over
`the 25% mark. This is the most reliable part of the recerved
`signal. Since the sample time is know (0.13), the nurnber of
`samples between these ‘transition points’ 1s used to calculate
`the period and hence the frequency. The frequency is then
`multiplied by 60 to compute the heart rate Gn bprn, beats per
`minute) of the patient.
`Is per divisson
`
`ayyy"
`16 waiiteeAisin
`
`Benod
`
`a0%
`
`Fig. 6. Heart rate measurement algorithm
`
`Fig. 5. Data packet composition
`
`5. Impact Algorithm
`
`is in beats per
`The heart rate included in the packet
`minutes (bpry which is calculated using the Heart Rate
`Algorithm explamed in Section V. The temperature data 1s
`
`the result ofthe ADC conversion, done once every 3 seconds,
`and is decoded into degrees Celsius at the receiver unit. The
`impact information is the maximum output voltage of the
`accelerometer measured in the past 3 seconds. This is also the
`
`digital value obtained from the ADC conversion.
`
`
`received from the accelerometer is a DC
`The signal
`voltage,
`the amplitude of which is directly related to the
`acceleration (62mV/g). This put is sampled at 10Hz, and
`stored in a buffer which is 10 bytes long. Every second, this
`buffer is scanned, and the maximum amplitude is recorded.
`Ad the recerrer unit,
`this data is compared to recorded
`representative values for walking and jogging. This ensures
`that an alarm is set off only for very sharp, heavy impacts.
`
`FITBIT, Ex. 1045
`
`FITBIT, Ex. 1045
`
`

`

`The reliability of impact detection was very good and the
`walking could easily be differentiated from a fall.
`
`with thumb on carotid artery). Figure 9 shows a segmentof a
`received heartrate signal.
`
`VI. PROTOTYPE AND EXPERIMENTAL RESULTS
`
`The Nordic micro-controller development board was used
`to build and test the prototype design. The analog processing
`circuitry and the sensors were assembled on PCBs which
`were placed within the wrist strap. Figure 7 shows the
`prototype hardware. The prototype was powered off a 9V
`battery. The RF transmission has been tested to operate
`successfully at 10 meters range through obstacles such as
`concrete walls. The receiver unit, without the casing, can be
`
`seen in figure 8.
`
`Fig, 8. Prototype receiver connected to a PC
`
`A. Heart Rate Sensor Test Results
`
`A new heart rate is calculated every 3.2 seconds. The
`output of the sensor is a 200-400 mV p-psignal. riding a DC
`signal of 5|00m¥V. On humantests it is able to measure within
`+2 bpm of a rested patient (confirmed by measuring pulse
`
`Fig. 9, Segmentof a heart rate signal
`
`B. Impact Sensor Test results
`
`The output of the accelerometer was tested with walking
`and simulated falling. The results showed the difference was
`simple to detect and proved the accuracy of the algorithm.
`Figure 10 shows the impact sensor output.
`
`
`
`Fig. 10, Impact sensor output for walking and a fall
`
`VIL DISCUSSIONS AND FUTURE DEVELPOMENTS
`
`In this paper we have presented the research, of applied
`nature, done to monitor physiological parameters such as skin
`temperature, heart rate and body impact. A prototype was
`successfully developed and tested to establish the proof of
`concept. The algorithms were tested and found to be accurate
`and reliable. The novel aspect of the design is its low cost and
`detection of medical distress which does not necessitate
`pressing any panic button. This is an enormous improvement
`over existing commercial products.
`An important aspect of the design was miniaturization, so
`that the system was as non-intrusive as possible to the wearer.
`This was achieved by the use of surface-mounted devices on
`the PCBs designed. Low power operational amplifiers were
`used to minimize battery consumption. The price of the unit
`currently is $70. The major costs come from the use of
`precision components, accelerometer and temperature sensor.
`With some modification, the system can be made available
`commercially, Future improvements will focus on the use of
`
`FITBIT, Ex. 1045
`
`FITBIT, Ex. 1045
`
`

`

`flexible PCBs to replace the stiff cards, so that it could be
`moulded around the wrist unit, making it more comfortable
`for the wearer.
`The design of the IR sensors could be improved to
`decrease its susceptibility to noise, to a point where it could
`be moved onto the wrist unit. This would provide a much
`more comfortable and less intrusive unit, getting rid of the
`finger glove.
`The addition of a blood-oxygen sensor would allow the
`system to detect medical distress more accurately by
`measuring the amount of oxygen in the blood (HbO). This
`could be implemented by the addition of another emitter
`diode operating at a wavelength which is more readily
`absorbed by oxygen, and measuring the light using photo
`detector. Blood pressure can also be measured by a technique
`known as ‘pulse delay’ [10] which involves calculating the
`time for the heart pulse to travel a known distance. This is
`directly related to the blood pressure of a patient, and allows
`for non-invasive measurement.
`The unit was initially designed for use by the elderly,
`within the house, where a caregiver is present but 1s not able
`to be constantly in visual contact with the subject. The
`receiver unit would be enhanced so that it can connect to
`
`either the local or cellular phone network, and in the case of
`an emergency would contact an ambulance. Beyond the
`application for elderly patients is the use by anyone who isat-
`risk, with a mental or physical disability. The device could be
`applied to prevent cot deaths in babies, by alerting the parents
`when the infant becomesstressed.
`
`2]
`
`3]
`
`A]
`
` 1]
`
`5]
`
`6]
`
`7]
`
`8]
`
`9]
`10 fee
`
`
`
`REFERENCES
`
`5S, Nakamoto H, Shinagawa Y, Tanikawa T., “A health
`Ohta
`monitoring system for elderly people living alone’,
`Joumal of
`Telemedicine and Telecare, Vol 8, No. 3, June 2002, pp. 151-156
`ee
`Dittmar
`A,
`Axisa
`F,
`Delhomme
`G,
`Gehin
`“New concepts and technologies in home care and ambulatory
`monitoring”, Studies in health technology and informatics, 2004, pp 9-
`35.
`Fazlur Rahman, Arun Kumar, G Nagendra, and Gourab Sen Gupta.
`"Network Approach for Physiological Parameter Measurement", IEEE
`Transactions on Instrumentation and Measurement, February 2005, Vol
`54, No.1, pp 337-346
`Jovanov E, Raskovic D, Price J, Chapman J, Moore A, Krishnamurthy
`A., “Patient Monitoring Using Personal Area Networks of Wireless
`Intelligent Sensors”, Biomedical Sciences Instrumentation, 2001, pp
`373-378
`Maki H, Yonezawa Y, Ogawa H, Sato H, Hahn AW, Caldwell WM.,
`“A welfare facility resident care support system”, Biomedical Sciences
`Instrumentation, 2004, pp 480-483.
`LifeLink
`Panic
`http:/weww.bgehome.com/hsprotection html#lifelink
`National Semiconductor, LM35 — Precision Centigrade Temperature
`Sensor, http://www national.com/pf/LM/LM35 html
`Y.C Sydney, A-Z Health Guide from WebMD: Medical tests, Body
`Temperature,
`http://www .webmd.com/hw/health_guide_atoz/hw198785.asp
`Eastern Michigan University, Lock-In Amplification
`http://www .physics.emich.edu/molab/lock-in‘index.html
`W.D Peterson., D.A.Skramsted, D.E-Glumac, Piezo Film Pulse
`Sensor,http://www.phoenix.te-iece.org/004_PiezoFilmBlood_
`FlowSensor/PhoenixPiezoPulse.htm
`
`overview,
`
`Button,
`
`FITBIT, Ex. 1045
`
`FITBIT, Ex. 1045
`
`