EXHIBIT 2133
`
`EXHIBIT 2133
`
`IPR2017—003 17
`CONDITIONAL MOTION TO AMEND
`
`VALENCELL, INC.
`EXHIBIT 2133 — PAGE 1
`
`IPR2017-00317
`CONDITIONAL MOTION TO AMEND
`
`VALENCELL, INC.
`EXHIBIT 2133 - PAGE 1
`
`

`

`lntemat‘ional Conference on Control. Automation and Systems 2007
`Oct. [WEI], 2007 in COEX, Seoul, Korea
`
`Development of a wearable health monitoring device with motion artifact reduced algorithm
`(ICCAS 2007)
`
`Hyonyoung Han. Yunjoo Lee and Jung Kim
`
`I Department of Mechanical Engineering. Korea Advanced Institute of Science and Technology- Daejeon- Korea
`(Tel : r82-4E-869-327l; E-mail: rangel@kaist.ac.kr)
`1 Department of Mechanical Engineering. Korea Advanced Institute of Science and Technology. Daejeolt. Korea
`(Tel : +82-42-869-3270: E-mail: rosela@kaist.ac.kr )
`3 Department of Mechanical Engineering. Korea Advanced Institute of Science and Technology. Daejcon, Korea
`( Tel : +8242-869-323 l; E-mai|:.lungkim@kaist.ac.kr ]
`
`Abstract: In this paper. a real-time. wearable and motion artifact reduced health monitoring device is represented. A
`finger band, wearable health monitoring device,
`is consists of phot‘oplethysmography {PPG} sensor, 3—axis
`accelerometer, microproeessor and wireless module. The PPG sensor acquire distorted heart beat signal due to motion
`artifact. The linger movements are detected using the accelerometer, and major motion directions causing of the noise
`are researched by comparing each directional motion signals and distorted PPG signal. Two directional motions are
`significantly related to noise.
`therefore.
`these two directional active noise cancellation algorithm was applied to
`reconstruct the noise added heart beat signal. Low order t4th order} NLMS (normalized Least Mean Square} adaptive
`filter is employed for small size wearable device. The finger band device is experimented in daily body motion
`condition (1—3 Hz). and reduce distortion rate less than 5‘! by active noise cancellation algorithm. The motion artifact
`reduced finger band sensor can offer continuous health monitoring without daily motion artifact.
`
`Keywords: real-time. wearable, photoplethysmography. motion artifact. active noise cancellation
`
`LINTRODUCTION
`
`Vital signal measurement devices and techniques are
`important research fie|d5 for health care monitoring and
`an emergency health alarm system for patients and the
`aged. These person who has Weak vital signal. should
`monitor their bio-signal, such as pulsation. continuously
`and alarm to others when their signal is weaker. Among
`various vital signals. pulsation is adequate to health
`monitoring device since it can be measured easier and it
`needs simple devices than others.
`Pltotopletltysmography (ping) is a noninvasive heart
`beat. pulsation. measurement
`instrument. which has a
`potential to he developed into a portable device due to
`its relatively small size. Heart beat affeci
`to blood
`pressm'e and it' change the volume of the vesset and rate
`of blood flow. These volume and flow change can be
`detected using near red and infrared wave length light
`source and detector. Using these properties. pulsation,
`which is related to heart heat. can be measured by
`gathering detector‘s signal.
`
`The instrument has been developed for. monitoring
`
`and diagl‘wsis patient not only in the hospital but also at
`home. However. for portable and ubiquitous health care
`
`To reduce the problem many methods were researched
`this hindrance; one is sensorless approaches, which
`
`“Sing
`frequency component
`“mm the pulsation‘s
`[2. 3]
`frequency and time domain Feature analysis
`without
`any other sensors. Another
`is
`the sensor
`approach. which obtain pulsation signal by removing
`the motion noises. using body movement
`information
`me other motion "icclt-IISIII'JI'I
`SEESGFS- The "1051
`well-known method 3'30“ this approach is active "OISE
`cancellation algorithm With an adaptive 5'19" [4? 5]- BL”
`most ot‘their research is otT—line analysis. large program
`size which is not adequate to wearable and portable
`device. and experiments are not mostly wireless system.
`This paper F‘WWMFi a real-time. wireless 311d wearable
`device with emotion artifact reduction algorithm. The
`device has an applicable 51113” pragl'amming size for
`portable devices.
`
`2- SYSTEM SETUP
`2'1 Hardware Description ,
`
`system, the real-time measurement, wireless and motion
`
`artifacts problems should be worked out.
`
`In particular.
`
`motion artifact reducnon '5 the most challenging issue.
`
`Fig. [Finger hand sensor
`The wearable sensor device should be small and light
`
`978-89-950038-6-2—9856tltt17tliIS 3'§_"|CR(IS
`IPR2017—003 17
`CONDITIONAL MOTION TO AMEND
`
`1 58 1
`
`VALENCELL, INC.
`EXHIBIT 2133 — PAGE 2
`
`IPR2017-00317
`CONDITIONAL MOTION TO AMEND
`
`VALENCELL, INC.
`EXHIBIT 2133 - PAGE 2
`
`

`

`
`
`
`
`and attach to body tightly to reduce noise effect and feel
`comfortable
`to wear. Additionally, wireless
`communication device also reduces motion artifact by
`reducing effect of the data cable inertia.
`The finger band sensor, wearable PPG sensor, is
`attached to the finger base, as shown in Fig. 1. The size
`and weight is 25 x 30 mm2 and 16.8g, respectively. The
`PPG sensor is located on the inner layer of the band, and
`the accelerometer is on the circuit.
`The finger band device is divided into two parts; the
`sensor device and the host analysis computer. In detail,
`the sensor device part can be divided into three parts, as
`shown in Fig.2.
`
`
`(cid:1659)
`
`Fig. 21 Block diagram of photoplethysmography
`
`
`The first sensor part consists of a light source, a photo
`sensor for PPG and an accelerometer for motion
`detection. A 940 nm wavelength, a surface mounted
`type Infrared LED, and a photo-diode are used to
`acquire the PPG signal. Although a more detailed
`explanation follow in the next chapter, motion is
`considered to be directly related to noise, so to measure
`and analysis the body motion and a noise source, a
`3-axis accelerometer is used.
`Next part is the pre-signal processing, circuit part,
`which contains an analog signal processing part,
`amplifiers, filters, and analog to digital converters
`(ADC). Since the raw signal on the sensor signal
`(especially the AC component) is so small and distorted,
`signal processing is demanded before being sent to a
`communication system. The raw signal demands a low
`pass filter for reducing high frequency noise and high
`pass filter for rejecting a DC component to enhance the
`AC component. As filters, second order active analog
`high and low pass filters (Sallen-Key Filter) are used.
`Filtering signals are amplified to enhance and acquire
`discriminable signals by a thousand times.
`The last part is digital signal processing. High order
`filtering has good performance to extract wanted signal,
`but more number of components are required to increase
`filtering order. Therefore, digital filtering is employed
`to satisfy both circuit size and filtering performance.
`The filters are designed as a 0.5 – 3 Hz band pass filter,
`
`and totally fourth order analog active filter and digital
`filter are used in this signal processing.
`The following part is the communication system. To
`transmit obtained data from the sensor device to the host
`computer,
`the microcontroller converts data
`into
`communication language, an 8bit digital signal. Then it
`transmits the data through a wireless device, Bluetooth.
`It is a widely used device, especially for portable
`devices such as in MP3 players or cellular phones. The
`device has up to 30m communication distance, and size
`of the device is 20 x 18 mm2. The operation voltage is
`3.7 V.
`Labview software from National Instrument obtains
`transmitted data and display by graph on the host
`computer. Also, digital filtering and analysis algorithm
`are programmed
`
`2.2. Motion artifact reduction algorithm
`Active noise cancellation algorithm is used to reduce
`motion artifact. This is a signal filtering algorithm from
`a noise added pulsation signal and body motion signal
`to noiseless pulsation signal. The distorted PPG signal
`contains motion signal components and using adaptive
`filter, heart beat signal is extracted from various noise
`signal components.
`(cid:1659)
`
`(cid:1659)
`
`Fig. 3. Block diagram of the active noise cancellation algorithm
`
`Fig. 3 shows a block diagram of an active noise
`cancellation algorithm, which reconstructs a raw
`pulsation signal (sk) from the corrupted signal (dk),
`using measurable noise signal (xk). Here, PPG and body
`motion data correspond to dk and xk respectively. This
`research predominantly used 3-axis accelerometer
`signals (xk) for body motion data (nk).
`In this study, NLMS (Normalized Least Mean
`Square) adaptive filters were employed due to their fast
`processing and low order filter coefficients [6]. In the
`equation (1), w(n), the digital filter coefficient is
`computed from products of step size ((cid:541)(n)), input data
`(x(n)) and error data (e(n)). Instead of fixed step size in
`LMS algorithm,
`the step size are changed and
`normalized by the energy of input data vector. Step size,
`
`1582
`
`IPR2017-00317
`CONDITIONAL MOTION TO AMEND
`
`VALENCELL, INC.
`EXHIBIT 2133 - PAGE 3
`
`

`

`almost same number of reference signal.
`
`
`Table 1 Error between signals on various
` frequency conditions
`(cid:1659) Z.C. error (%)
`1.5 (Hz)
`2.0 (Hz)
`2.5 (Hz)
`
`25.00
`
`57.89
`
`74.36
`
`2.78
`
`5.26
`
`2.56
`
`Signal
`comparison
`Reference v.s.
`Measurement
`Reference v.s.
`Reconstructed
`
`
`Table 1 shows error rates between the numbers of
`each signal’s zero-crossing
`in various
`frequency
`conditions. Error rates increase from 25 % to 75 % as
`the frequency is higher. However reconstructed PPG
`signal case have low error rate, averagely lower than 5%.
`Besides it is represented lower than 4 count on beat per
`minute unit. Especially, in the 2.5 Hz motion condition,
`the error rate between reference and measured signal is
`74 %, but between reference and reconstructed signal is
`2.56 %. It means that the designed device and algorithm
`reduce motion artifacts efficiently.
`
`
`4. CONCLUSION
`
`ref.
`
`corrupted
`
`reconstructed
`
`
`
`4
`
`0
`
`0
`
`5
`
`10
`
`15
`
`Time(sec)
`Fig.5. Reference (bottom), corrupted (middle) and reconstructed (top)
`signal in 2.5 Hz hand waving conditions
`
`
`
` A
`
` real-time, wearable and wireless finger band sensor
`is designed with
`the motion artifact
`reduction
`algorithms. We obtain body motion data which is source
`of motion artifact using the accelerometer, and it is
`applied to active noise cancellation algorithm. As
`experiments, one directional hand motions which has
`difference frequency conditions are given. Fig.5 shows
`experimental signals;
`lower periodical signal
`is
`reference pulsation signal, middle complex signal is
`measured corrupted signal, and
`top smooth and
`periodical signal is reconstructed signal. On the graph
`also represent the performance is well-done. As a result,
`counting error of pulsation signal is reduced less than
`
`1583
`
`
`
`(1)
`
`(2)
`
`3. EXPERIMENT
`PPG signals are measured at not only moving left
`finger but also right finger in fixed pose, as a reference
`signal. Pre-experiment resulted that longitudinal axis
`and rotational direction to finger directional movements
`are strongly related to motion artifact. Therefore, two
`directional active noise cancellation algorithms are
`applied to reconstruct the pure PPG signal. The
`experiments are progressed in that various frequency
`movement conditions are experimented during 30 sec in
`finger longitudinally and 20-cm-long hand waving. The
`range of the frequency is similar to daily body motion,
`1.5~2.5 Hz. Zero-crossing (Z.C) peak counting method
`are used as an evaluation method.
`
`
`(cid:52)(cid:71)(cid:72)(cid:71)(cid:84)(cid:71)(cid:80)(cid:69)(cid:71)(cid:3)(cid:50)(cid:50)(cid:41) (cid:47)(cid:71)(cid:67)(cid:85)(cid:87)(cid:84)(cid:71)(cid:70)(cid:3)(cid:50)(cid:50)(cid:41) (cid:52)(cid:71)(cid:69)(cid:81)(cid:80)(cid:85)(cid:86)(cid:84)(cid:87)(cid:69)(cid:86)(cid:71)(cid:70)(cid:3)(cid:50)(cid:50)(cid:41)
`
`(cid:19)(cid:20)(cid:18)
`
`(cid:19)(cid:21)(cid:24)
`
`(cid:27)(cid:18)
`
`(cid:25)(cid:22)
`
`(cid:25)(cid:20)
`
`(cid:25)(cid:24)
`
`(cid:25)(cid:20)
`
`(cid:25)(cid:26)
`
`(cid:25)(cid:24)
`
`(cid:19)(cid:24)(cid:18)
`(cid:19)(cid:22)(cid:18)
`(cid:19)(cid:20)(cid:18)
`(cid:19)(cid:18)(cid:18)
`(cid:26)(cid:18)
`(cid:24)(cid:18)
`(cid:22)(cid:18)
`(cid:20)(cid:18)
`(cid:18)
`
`(cid:60)(cid:71)(cid:84)(cid:81)(cid:15)(cid:37)(cid:84)(cid:81)(cid:85)(cid:85)(cid:75)(cid:80)(cid:73)
`
`(cid:19)(cid:16)(cid:23)
`
`(cid:20)(cid:16)(cid:23)
`
`(cid:1659)
`
`(cid:20)(cid:16)(cid:18)
`(cid:40)(cid:84)(cid:71)(cid:83)(cid:87)(cid:71)(cid:80)(cid:69)(cid:91)(cid:3)(cid:10)(cid:42)(cid:92)(cid:11)
`Fig.4. Pulsation counts from the Z.C. of the signals
`
`
`
`The result, Fig.4, shows that the motion artifact
`reduction algorithm improve performance. The x, y axis
`represent
`frequency
`and
`represent number of
`zero-crossing during a minute, respectively. Lower than
`1Hz condition, zero-crossing error between reference
`and distorted measured PPG signal are nearly zero. The
`number of reference pulsation is 72 to 78 bpm (beat per
`minute), from normal healthy subject, and That of
`measured signal is 90 to 136 bpm in 1.5~2.5 Hz hand
`waving conditions, the error rates increase as movement
`frequency is higher. However, that of the reconstructed
`signal with motion artifact reduced algorithm detected
`
`(cid:541), are computed with the coefficients a, b and input data
`as equation (2), and the role of the coefficients are
`prevent the step size not to fluctuate excessively.
`Thorough this various step size condition, in NLMS
`cases, more flexible and stable signal processing is
`possible, which is appropriate for real-time and wireless
`sensors.
`w
`(
`
`n
`
`1)
`(cid:14) (cid:32)
`
`w
`
`
`
`( )n
`
`(cid:16)
`
`(cid:541)
`
`
`
`( )n
`
`(cid:32)
`
`a
`
`
`
`
`
`
`
`(cid:541)
`
`( ) ( ) ( )n n e nx
`b
`x x
`(cid:14) T
`
`IPR2017-00317
`CONDITIONAL MOTION TO AMEND
`
`VALENCELL, INC.
`EXHIBIT 2133 - PAGE 4
`
`

`

`
`
`
`
`5%.
`The experimental frequency condition is similar to a
`hand’s daily movement. For example as hand motions,
`running, walking [7], and hand gesture [8] have
`approximately 2 to 4 Hz frequency motion, and as
`finger motions, object exploration and texture scanning
`[9] have 0.8 to 2 Hz in table 2. But in hand writing,
`typing and tapping condition (4~7 Hz), they demand
`more high frequency artifact experiments and it remains
`future works.
`
`
`Table 2 Frequency of daily hand movement
`Motion
`Frequency (Hz)
`3
`2
`0 - 4
`0.8 – 1.6
`0.8 – 2.2
`
`Hand motion
`
`Finger motion
`
`4 - 7
`
`Run
`Walk
`Hand gesture
`Object
`exploration
`Texture scan
`Hand writing
`Typing
`Tapping
`
`variable step-size LMS filter," 2002.
`[5] H. Harry Asada, H. H. Jiang, and P. Gibbs,
`"Active noise cancellation using MEMS
`accelerometers for motion-tolerant wearable
`bio-sensors," Conf Proc IEEE Eng Med Biol
`Soc, vol. 3, pp. 2157-60, 2004.
`[6] S. M. K. a. D. Morgan, Active Noise Control
`Systems: Algorithms and DSP Implementations:
`JOHN WILEY \& Sons, Inc., 1995.
`[7] Akiko OGAWA, Yusuke KONISHI, Ryosuke
`SHIBAZAKI, “Identification of human activity
`modes with wearable sensors for autonomous
`human positioning system,”, Geoinformation
`forum Japan 2002, Student forum, 2002.
`[8] Y. X. a. F. Quek, "Hand Motion Gesture
`Frequency Properties and Multimodal Discouse
`Analysis," International Journal of Computer
`Vision, vol. 69, pp. 353-371, 2006.
`[9] E. Kunesch, F. Binkofski, and H. J. Freund,
`"Invariant
`temporal
`characteristics
`of
`manipulative hand movements," Exp Brain Res,
`vol. 78, pp. 539-46, 1989.
`
`
`Humans are moved more than remain fixed, relaxed
`pose on our daily life. Moreover, accidents and
`emergencies occur in movement condition. Therefore,
`wearable and portable health care device with motion
`artifact reduced algorithm techniques are demanded for
`continuous health monitoring. And the proposed device
`in this paper could have a potential to developing this
`portable and wearable device without daily motion
`artifact.
`
`
`ACKNOWLEDGMENT
`
`
`This work was supported by the IT R&D program of
`MIC/IITA. [2005-S-096-02, Development of wearable
`PC interface technology for the disabled ]
`
`
`REFERENCES
`
`
`J.G.Webster, Design of Pulse Oximeters, 1997.
`[1]
`[2] B. S. Kim and S. K. Yoo, "Motion artifact
`reduction
`in photoplethysmography using
`independent component analysis," IEEE Trans
`Biomed Eng, vol. 53, pp. 566-8, 2006.
`[3] Y. S. Yan, C. C. Poon, and Y. T. Zhang,
`"Reduction of motion artifact in pulse oximetry
`by smoothed pseudo Wigner-Ville distribution,"
`J Neuroengineering Rehabil, vol. 2, pp. 3, 2005.
`[4] K. W. Chan and Y. T. Zhang, "Adaptive
`reduction
`of motion
`artifact
`from
`photoplethysmographic
`recordings using a
`
`1584
`
`IPR2017-00317
`CONDITIONAL MOTION TO AMEND
`
`VALENCELL, INC.
`EXHIBIT 2133 - PAGE 5
`
`