International Journal of Scientific Research and Engineering Development-– Volume 1 Issue 2, Nov-Dec2018

Available at www.ijsred.com
RESEARCH ARTICLE OPEN ACCESS

The Comparison Test of Respiratory Rate

Changjian Deng*
*(Department of Control Engineering, Chengdu University of Information Technology, and Chengdu,China)
----------------------------------------************************----------------------------------
Abstract:
The body respiratory rate (RR) is the optional parameters of vital signal, and it is one of the monitoring
indicators of physiological distress. The paper presents a test method of body respiratory rate by using the
status sensors and infrared imaging sensors. The method of the Kalman filter and zero crossing test are
applied in the RR test, the measurement and simulation results show its feasibility. And they are compared
with testing by commercial pressure sensors module, the results show the effective of methods

Keywords —Body Respiratory Rate , Status Sensors , Kalman Filter, Feature Extraction.
----------------------------------------************************----------------------------------
The accelerometers are often used in stationary
I. INTRODUCTION test, for the movement or cline of the body may
There are two kinds of respiratory rate(RR) test change the value and make the results untrusty. The
methods, one is using contact sensor, another is other test methods include infrared imaging sensors,
using non-contact sensor. The contact RR test often the capacity sensor, and the pressure sensors.
adopt the electrical impedance sensor,the pressure Nadine Hochhausen presents a robust and effective
sensor, the movement sensor,the ECG and PPG algorithm for BR detection in thermal videos, it
sensors; the non-contact test uses the infrared detects the movement of the nose[6]; David
imaging sensors, the capacity sensor, Fiber Bragg Naranjo-Hernández and so on discusse a method
Grating sensor ,and so on. based on the measurement of the capacitance
In [1],AndreaAliverti reviews the currently existing between two electrodes[7]. They use four-
technology in respiratory health and disease, for step algorithm:1) first ,use the low-pass filter to
example, the monitoring of sleep disorders for decrease the noise ; 2) peak values are calculated to
timely diagnosis and treatment. Therespiratory get the time instants;3)the calculated period instants
inductive plethysmography is hard to monitor is considered, and to correct the time instants
during the daily life. Accelerometers are used to number;4) the respiratory cycle and respiratory rate
derive breathing rate by measuring movements of are estimated. These methods are also used in
the chest wall[2][3]. pressure sensor testing, for their signal is almost
The principle of test is that MEMS same[8].
accelerometers can measure inclination changes The current infrared imaging test is not suitable
during breathing, and to obtain a respiratory rate. for smart vest, and the air sacs, air tube, and band
The algorithms to test RR include Fourier make the pressure sensor also not suitable for smart
analysis[4], adaptive filter, Principal Component vest design.
Analysis[5], and so on. For example, Spire In the paper, the infrared imaging sensor and the
(www.spire.io) and the Vitali Smart Bra&GEM state sensor embedded in smart vest are considered,
(https://vitaliwear.com), theyanalyze breathing the test results are compared with commercial
patterns in terms of breathing frequency and pressured sensors.
waveform. The rest of the paper is organized as follows.
Section 2 describes the material and methods to test

ISSN: 2581-7175 ©IJSRED: All Rights are Reserved Page 192

International Journal of Scientific Research and Engineering Development-– Volume 1 Issue 2, Nov-Dec2018
Available at www.ijsred.com
the respiratory rate. Section 3 presents the main 3)pressure sensors
results, and Section 4 gives the conclusions. pressure sensor is a device for pressure
measurement of gases or liquids. Pressure is an
II. MATERIAL AND METHODS expression of the force required to stop a fluid from
An easy way to comply with the conference expanding and is usually stated in terms of force per
paper formatting requirements is to use this unit area. A pressure sensor usually acts as a
document as a template and simply type your text transducer; it generates a signal as a function of the
into it. pressure imposed. For the purposes of this article,
A. Methodology for Respiratory Sensing such a signal is electrical.
Pressure sensors are used for control and
1)State sensor
monitoring in thousands of everyday applications.
The MPU-6050 devices combine a 3-axis
Pressure sensors can also be used to indirectly
gyroscope and a 3-axis accelerometer on the same
measure other variables such as fluid/gas flow,
silicon die, together with an onboard Digital Motion
speed, water level, and altitude. Pressure sensors
Processor™ (DMP™), which processes complex 6-
can alternatively be called pressure transducers,
axis Motion Fusion algorithms. The device can
pressure transmitters, pressure senders, pressure
access external magnetometers or other sensors
indicators, piezometers and manometers, among
through an auxiliary master I²C bus, allowing the
other names.
devices to gather a full set of sensor data without
Pressure sensors can vary drastically in
intervention from the system processor.
technology, design, performance, application
The MPU-60X0 features three 16-bit analog-to-
suitability and cost. A conservative estimate would
digital converters (ADCs) for digitizing the
be that there may be over 50 technologies and at
gyroscope outputs and three 16-bit ADCs for
least 300 companies making pressure sensors
digitizing the accelerometer outputs.
worldwide.
2) infrared imaging sensor
Uses the piezoresistive effect of bonded or
The Infrared Array Sensor is a thermopile-typed
formed strain gauges to detect strain due to applied
infrared sensor which detects quantity of infrared
pressure, resistance increasing as pressure deforms
ray. Generally, temperature accuracy will be
the material. Common technology types are Silicon
degraded in the following situations. Be sure to
(Monocrystalline), Polysilicon Thin Film, Bonded
verify performance and reliability under actual
Metal Foil, Thick Film, Silicon-on-Sapphire and
conditions of use and make any necessary
Sputtered Thin Film. Generally, the strain gauges
temperature corrections.
are connected to form a Wheatstone bridge circuit
・There is a heat emitting body located close to to maximize the output of the sensor and to reduce
where the sensor is mounted. sensitivity to errors. This is the most commonly
・A flow of warm or cold air is hitting the sensor. employed sensing technology for general purpose
pressure measurement.
・ The temperature of the sensor is subject to
B. Methodology for Signal processing
sudden change.
1) quaternion-based and kalman filter for
・When an object made of glass, acrylic or other MPU-60X0
subject which far infrared rays have difficult The paper compares the quaternion-based filter
passing through is located between the sensor and and Kalman filter.
what is to be detected. Kris Winer presents a quaternion-based filter that
・A substance (dirt or water droplets) that makes can be used in Arduino-mini [9],the procedure of
it difficult for far infrared rays to pass through is program is in below:
attached to the sensor lens step1: Auxiliary variables to avoid repeated
arithmetic;

ISSN: 2581-7175 ©IJSRED: All Rights are Reserved Page 193

International Journal of Scientific Research and Engineering Development-– Volume 1 Issue 2, Nov-Dec2018
Available at www.ijsred.com
+ − −
( )= 2( + ) "
step2: Normalize accelerometer measurement

2( − )
step3: Compute the objective function and

2( − ) 2( + )
Jacobian

" 2( − )
step4: Compute the gradient (matrix
− + − #
multiplication)
2( + ) − − +
step5: Normalize the gradient

*
step6: Compute estimated gyroscope biases
0
(3)

Use ( $%% ) &0) = &* ),obtain

step7: Compute and remove gyroscope biases

*
step8: Compute the quaternion derivative
1
2( $%% $%% + $%% $%% ) = *
step9: Compute then integrate estimated

+ 2( $%% $%% − $%% $%% ) = * ",

quaternion derivative
step10: Normalize the quaternion

$%% − $%% − $%% + $%% = *

Kristian Lauszus provides a Kalman filter library (4)
for Arduino [10]. The procedure of program is at
below: The principle of kalman filter is described as

,- = .- /-01 + 2- 3- + 4-
Step1: set the variables; below[12]:

.- , the state-transition model;

Step2: Reset bias; (5)

5- , the observation model;

Step3: Discrete Kalman filter time update

6- , the covariance of the process noise;

equations - Time Update ("Predict");

7 - , the covariance of the observation noise;

A) Updatexhat - Project the state ahead

2- , the control-input model;

B) Update estimation error covariance - Project

.- ,is the state transition model which is applied

the error covariance ahead

to the previous state ,-01 ,;

Step4: Discrete Kalman filter measurement

8- , is the process noise which is assumed to be

update equations - Measurement Update;
A) Calculate Kalman gain - Compute the Kalman

distribution, , with covariance, 4- ~:(;, =- ).

gain; drawn from a zero mean multivariate normal

At time k an observation (or measurement) >- of

B) Calculate angle and bias - Update estimate

the true state ,- is made according to

with measurement zk;

>- = 5- /- + ?-
Step5: Calculate estimation error covariance -

5- is the observation model which maps the true

Update the error covariance. (6)

state space into the observed space and @- is the

The principle of quaternion-based is described as
below[11]:

= = mean Gaussian white noise with covariance 7 - :

a unit quaternion observation noise which is assumed to be zero

(1) @- ~:(;, - ）
where α is the rotation angle and e is the unit Predict:

A-|-01 = .- /
/ A-01|-01 + 2- 3- (7)
vector that represents the rotation axis. State estimate:
quaternion multiplication, given two quaternions

− − − C-|-01 = .- C-01|-01 DE- + =- (8)

p and q, is defined as Error covariance:

+ + −
⨂ = − + +
Update:

+ − +
(2)
FG- = >- − 5- /
A-|-01
Innovation or measurement prefit residual:

( ) is the direct cosine matrix (DCM) given in

(9)

H- = - + 5- C-|-01 IE-
Innovation covariance:
terms of the orientation quaternion. (10)

ISSN: 2581-7175 ©IJSRED: All Rights are Reserved Page 194

International Journal of Scientific Research and Engineering Development-– Volume 1 Issue 2, Nov-Dec2018
Available at www.ijsred.com

J- = C-|-01 IE- H01

Optimal Kalman gain: even statistical operators. For example, Gaussian
- (11) low-pass, adaptive median filter, and MSR method.

A-|- = /
/ A-|-01 + K- LM-
Updated state estimate: The MSR performed well on enhancing detail
(12) and removal of uneven-heating compared to other

C-|- = (N − J- I- )C-|-01 (N − J- I- )E +
Updated state covariance: photometric normalization techniques. This method
is seen to resemble how our eyes (retina) perceive
J- - JE- (13)
variations in grayscale intensities.
B) Thermal Contrast Enhancement
M-|- = >- − 5- /
L A-|-
Measurement post-fit residual: Absolute contrast is the classical and most basic
(14) contrast definition describing the excess

OPQR(S) = T(U) − THP (S)

2) The zero-crossing test temperature over a defect-free region:
In a sine wave or other simple waveform, this (15)
normally occurs twice during each cycle.It is a where T(t) is the temperature for any pixel and
device for detecting the point where the voltage TSa(t) is the temperature for chosen sound area at
crosses zero in either direction. The algorithm is time t.
used in state sensor to calculate the breathing rate Standard contrast was developed to suppress the
[13]. impact of reflections produced by the surrounding
The zero-crossing rate(ZCR) is the rate of sign- environment through subtraction of an image's
changes along a signal, i.e., the rate at which the
ORSV (S) = T
E(S)0E(S; )
pixels at time t0 before heating.
signal changes from positive to negative or back.
HP (S)0THP (S; )
Step1: A DC offset is computed and subtracted. (16)
Step2: The ZCR is then computed, focusing on C)Data Normalization
the changes of sign of the signal, using a sample- The temperature can be expressed as an excess
by-sample sequential algorithm: surface temperature measurement that indicates a
Step3: The Respiratory rate is calculated. change in surface temperature after the heat has
The infrared imaging change algorithm been applied to the target sample. To display the
The principle of respiratory cycle test is: during excess temperature as a grayscale image, data
inspiration, cold air from the environment is inhaled, normalization needs to be performed.
and during expiration, warm air from the lungs is
exhaled. thermal imaging detects this temperature III. RESULTS
modulation to test respiratory rate. The above three methods is experimented in a
Step1: region selection and tracking; smart vest, the smart vest uses the sensor module to
Step2: extraction of the respiratory waveform test body respiratory rate, the comparison of these
Step3: statistical Analysis three methods is described in Table.1
The Thermographic Image Processing methods
include: Thermal Contrast Enhancement, Denoise TABLE 1. THE COMPARISON OF THREE RESPIRATORY RATE METHODS IN
SMART VEST
method, Data Normalization, and so on.
A) Denoise method State
pressure
infrared
sensor imaging
The thermal signal has a spatial resolution of Nx module
sensors module
sensor
× Ny pixels. To handle the thermal sequence, the
convenience Good Bad normal
frames can be concatenated as a three-dimensional
Nx × Ny × Nt matrix, where Nt de fines the total precise normal Good normal
number of frames in the sequence. thickness Good Bad normal
Thermograms are prone to ambient and inherent
camera noise. There are many ways to de-speckle Weight Good Bad normal
or de-noise an image using linear, non-linear or

ISSN: 2581-7175 ©IJSRED: All Rights are Reserved Page 195

International Journal of Scientific Research and Engineering Development-– Volume 1 Issue 2, Nov-Dec2018
Available at www.ijsred.com
A State sensor module The correction of RR: 3 times/ per minute;
Paper selects the MPU6050 state sensor, the The test results is output to PC through serial port.
experiment platform is shown in Fig.1.
C The infrared imaging sensor module
The paper selects AMG8833 sensor module, its
experiment platform is shown in Fig.3.

Fig.3 the infrared imaging sensor experiment platform

Fig.1 the state sensor experiment platform

The test data is output to a PC, the Matlab is used
to verify the algorithms.
The program has three stages:
Stage 1: open the serial port, and acquire the state
data;
Stage 2: pre-process the data and plot its raw
figure; include quaternion-based and kalman filter
Stage 3: zero-crossing test and the respiratory rate
calculation.
The test data is output to a PC, the MATLAB is
used to analyse the data.
The program includes three stage:
Stage 1: open the serial port, and acquire the
pixel image data;
Stage 2: pre-process and plot its raw figure; it is
shown in Fig.4.
Stage 3: Data Normalization, Thermal Contrast
Enhancement and the respiratory rate calculation.

B The pressure sensors module 1

The paper selects the DCLK365 module to test 2

RR. The platform is shown in Fig.2. 3

1 2 3 4 5 6 7 8

Fig.4. The infrared imaging of AMG8833

Fig.2 the pressure sensors module experiment platform D The test results
The operation condition: With the test platform and installation in the
Temperature: 5°C-40°C smart vest, paper tests many times of every sensor
Relative humidity: 18%-80% modules, the test man is same one, during sleep or
Air pressure: 80kPa-105kPa set down in a chair. The results are shown in
Power voltage:3.3V-5V Table.2.
The Test range:
0~39.9kPa;
Resolution: 0.1kPa,Precision:0.4kPa;

ISSN: 2581-7175 ©IJSRED: All Rights are Reserved Page 196

International Journal of Scientific Research and Engineering Development-– Volume 1 Issue 2, Nov-Dec2018
Available at www.ijsred.com
TABLE 2. THE TEST RESULTS OF THREE RESPIRATORY RATE METHODS accelerometer data. In: 2010 International
Average Average Conference on Body Sensor Networks (BSN). New
Average value
value
(pressure
value York, IEEE, 2010; pp. 144–150.
(State (infrared [4] T. Reinvuo, M. Hannula, and H. Sorvoja,
sensors
sensor imaging
module)
module)
sensor)
“Measurement of respiratory rate with high-
Sleep (test resolution accelerometer and EMFit pressure
21 18 25
platform) sensor,” Proc. IEEE Sensors Appl. Symp., pp. 192–
set down in a 195, Jan 2006.
chair (test 24 19 27 [5] C. Davis, A. Mazzolini, and D. Murphy, “A new
platform)
Sleep (smart fibre optic sensor for respiratory monitoring,”
22 19 29 Australas. Phys. Eng. Sci. Med., vol. 20, no. 4, pp.
vest)
set down in a 214–9, Dec 1997.
chair (smart 25 19 27 [6] Hochhausen N , Pereira C B , Leonhardt S , et al.
vest)
Estimating Respiratory Rate in Post-Anesthesia
IV. CONCLUSIONS Care Unit Patients Using Infrared Thermography:
An Observational Study[J]. Sensors, 2018, 18(5).
The paper uses MPU6050 state sensor, [7] Naranjohernández D, Talaminosbarroso A,
AMG8833 infrared imaging sensor module and Reinatosina J, et al. Smart Vest for Respiratory
DCLK365 pressure sensors module to test the Rate Monitoring of COPD Patients Based on Non-
respiratory rate. Contact Capacitive Sensing[J]. Sensors, 2018,
The algorithms used in MPU6050 state sensor 18(7).
module include quaternion-based filter algorithms, [8] Nizami, S.; Bekele, A.; Hozayen, M.; Greenwood,
kalman filter algorithms, and zero-crossing test K.; Harrold, J.; Green, J. Comparing time and
algorithms. frequency domain estimation of neonatal
The algorithms used in AMG8833 infrared respiratory rate using pressure-sensitivemats. In
Proceedings of the 2017 IEEE International
imaging sensor module include Data Normalization
Symposium on Medical Measurements and
algorithm, Thermal Contrast Enhancement Applications, MeMeA 2017, Rochester, MN, USA,
algorithm and the respiratory rate calculation 7–10 May 2017; pp. 239–244.
algorithm. [9] K. Winer, T 2014 Invensense Developer's
The experiment test in test platform and smart Conference,acess 2018.12.11, availabe:
vest show the effective of the design methods. https://github.com/kriswiner/MPU6050/wiki/2014-
Invensense-Developer%27s-Conference
ACKNOWLEDGMENT [10] Kristian Lauszus, TKJ Electronics 2012,
This works is supported by Department of https://github.com/TKJElectronics/KalmanFilter
Control Engineering, Chengdu University of [11] Yun X, Bachmann E R. Design, Implementation,
Information Technology. and Experimental Results of a Quaternion-Based
Kalman Filter for Human Body Motion Tracking[J].
REFERENCES IEEE Transactions on Robotics, 2006, 22(6):1216-
[1] Aliverti,A. Wearable technology: Role in 1227.
respiratory health and disease. Breathe 2017, 13, [12] Feng K, Li J, Zhang X, et al. An Improved Strong
e27–e36 Tracking Cubature Kalman Filter for GPS/INS
[2] Hung PD, Bonnet S, Guillemaud R, et al. Integrated Navigation Systems.[J]. Sensors, 2018,
Estimation of respiratory waveform using an 18(6):1919.
accelerometer. In: 5th IEEE International [13] Rizqyawan M I , Simbolon A I , Suhendra M A , et
Symposium on Biomedical Imaging: From Nano to al. Mobile/android application for QRS detection
Macro, 2008. New York, IEEE, 2008; pp. 1493– using zero cross method[J]. Journal of Physics
1496. Conference Series, 2018, 978.
[3] Bates A, Ling MJ, Mann J, et al. Respiratory rate
and flow waveform estimation from tri-axial

ISSN: 2581-7175 ©IJSRED: All Rights are Reserved Page 197