2017 International Conference on Security, Pattern Analysis,

and Cybernetics (SPAC)

Reflective type blood oxygen saturation detection

system based on MAX30100
Jiaxi Wan1, Yuhua Zou2
Ye Li*, Jun Wang2
School of Electronic Engineering and Automation, Guilin
Shenzhen Institutes of Advanced Technology, Chinese
University of Electronic Technology
Academy of Sciences
Guilin, China
Shenzhen, China
Email: wjx_funny24@163.com
Email: *ye.li@siat.ac.cn

Abstract—Oxygen is the key material to human’s life, and detection as shown in Fig.1(b), two light tube and a receiving
oxygen saturation is one of the important indexes reflecting
tube in the detected parts of the same side. So this type of
organic oxygen delivery status. Determining oxygen saturation in
human blood by transmittance oxymetry is well developed as a sensor can be applied anywhere on the body surface. The
monitoring technique, which is applied widely in clinical Photodiode receives the reflected light from the surface, and
diagnosis or house health care. However, reflectance detecting is oxygen saturation is calculated by different intensity of the
necessary in many circumstances, such as cerebral oxygen reflected light received.
saturation, muscle oxygen saturation, or fetal oxygen saturation
monitoring. So the reflective type oxygen saturation detection 3KRWRGLRGH
system gradually has become main flow of development.
Based on the principle of oxygen saturation measurement, this
paper introduces a blood oxygen saturation detection system
design scheme based on the integrated chip MAX30100, which
can simplify the circuit design, reduce system footprint, reduce
the designing time and system power consumption. Through 5HG /(' ,5 /('
introduces the system hardware and software structure, signal
(D)
processing methods and other aspects of the study to achieve the
fingertip pulse signal acquisition and noise reduction processing.
Through practical test, the system prototype machine realizes the
function of pulse oxygen saturation detection.

Key words: reflective type, SpO2, ultra-low cost, MAX30100 5HG /(' ,5 /(' 3KRWRGLRGH
(E)
Fig.1: the structure of the oxygen saturation detection
I. INTRODUCTION
Blood oxygen saturation, is the concentration of blood II. BASIC THEORY
oxygen in the blood, which is an important physiological
Non-invasive oxygen saturation measurement is based on
parameter of the respiratory and circulatory system. Hypoxia
the principle that arterial blood absorbs light as the arterial
can lead to inattention, memory loss, dizziness, anxiety and
pulse changes, and the pulse wave pulsation state is reflected
other symptoms; myocardial hypoxia is very sensitive to acute by the change of light absorption amount. According to the
myocardial hypoxia, there may be ventricular fibrillation and Beer-Lambert law, when the arterial pulsation, the changes of
cardiac arrest; severe or long-term hypoxia, leading to volume will make the light path transmission of light changes,
myocardial failure, blood pressure, blood circulation failure; and the light absorption of arterial blood changes, resulting in
what is worse, hypoxia directly damage the cerebral cortex, changes in the intensity of light received by the photodiode [4].
resulting in brain tissue degeneration and necrosis[1]. Therefore, the oxygen saturation can be obtained by this
However, lots of people always have high pressure from study change signal.
and work in modern society. They don’t have enough time to
go to hospital, so convenient and non-invasive device is an I = IOe–kcd (1)
important study direction in increase blood oxygen saturation. In the equation (1), (2) and (3), the IO is a monochromatic
At present, the majority of blood oxygen monitor use light intensity, the c, c1 and c2 is the blood concentration, the I
transmission detecting method [2], that is, the detected parts is the intensity of light reflected back, the k0, k1 and k2 is the
(such as fingers) placed between the LED and receiver tube[3- light absorption coefficient, and the d is the optical path.
4]. The structure of transmission type oxygen saturation
detection is shown in Fig.1(a). However, because of the I = IOe–K0C0d0 e–K1C1d1 e–K2C2d2 (2)
transmission type oxygen sensor using a limited range, can’t
use the transmission type oxygen sensor in the body parts ∆I = I − IO = IOe–K0C0de–K1C1(d–∆d)e–K2C2(d–∆d) (3)
(such as the forehead, chest, etc.) to detect oxygen saturation,
The equation (4), ∆I stands for AC signal, I stands for
and measurement for a long time. Not only make the surveyor DC signal, and I is much larger than ∆I.
feel uncomfortable, but also influence the accuracy of the
IAC
measurement results. Different from the transmission type = ∆I = −(k1c1 + k2c2)∆d (4)
IDC I
oxygen saturation detection, reflectance oxygen saturation

978-1-5386-3016-7/17/$31.00 ©2017 IEEE 1

 %
+E2²
+E pulsatile $&

$PSOLWXGH
&RPSRQHQW
$EVRUSWLRQ

+E
'&
Non-pulsatile &RPSRQHQW
+E2²
660 810 940 QP 7LPH
: Fig.3: The AC and DC component of PPG single
DYHOHQJWK According to research on oxygen saturation field, the
Fig.2: Absorption of HbO2 and Hb at Different Wavelengths relationship between SpO2 and different oxygenated level is
The SpO2 is defined as the ratio of the level oxygenated not linear [7]. So we can define a coefficient R.
hemoglobin (HbO2) over the total hemoglobin (Hb) level [5]: K1Hb0 CHb0 +K1HbCHb
Ifi1 /Ifi1
AC DC 2 2 Sp02 ∝ R (6)
c R = =
Hb02 K2Hb0 CHb0 +K2HbCHb
Ifi2 /Ifi2
Sp02 = cHb02+cHb × 100% (5) AC DC 2 2

We can conclude that HbO2 and Hb have significant Two different wavelength are h1 and h2. Light transmit
differences in the absorption coefficients of red and infrared through tissue and blood vessels will be divided into DC
light. As shown in Fig.2 [6], the absorption coefficient of Hb component (such as skin, muscle, bone, etc.) and AC
in the red spectrum region (600nm-700nm) is much larger than component (such as arterial blood).
that of HbO2, the absorption coefficient of Hb in the infrared Because of the differences among individuals, we usually
spectrum region (810nm-1000nm) is smaller than that of use empirical formula to calculate the blood oxygen saturation
HbO2. Therefore, the system uses a light source of 660nm red [7-10].
and 940nm infrared light.
Sp02 = AR2 + BR + C (7)
Where A, B and C can be calculate by experiment.

32:(5 028'/(
3.39 1.89 3.39

0$;30100 )/$6+ &211(&725

$0%,(17 /,*+7
&$1&(//$7,21 ',*,7$/
* 3
',*,7$/'$7$
$'& ),/7(55(*,67(5
, 2
&RUWH[–03 %/8(7227+ 02'8/(
$1$/2* ,,& ,,&
5(' ,5
3+272
',2'(

26&,//$725 /(' '5,9(56 86$57 &211(&725

0DVWHU &RQWUROOHU

Fig.4 The hardware overall framework of system

saves the volume of the entire system to improve the degree of
III. SYSTEM ARCHITECTURE acquisition and pretreatment function, but also greatly

A. Hardware
In order to achieve the miniaturization and wearable of
the detection system, the system hardware uses a large number
of integrated circuit chips and is mainly divided into four parts:
power module, analog front end, center microprocessor and
Bluetooth module. The hardware overall framework of system
is shown as Fig.4.
Considering portability and stability, we choose the ultra-
low power MAX30100 chip produced by Maxim Integrated
Company. This analog front end integrates the necessary
function modules such as pulse control circuit, filter amplifier
circuit, LED and LED driver circuit and A/D conversion
module, thus replacing the traditional complex peripheral
circuit design. Not only to achieve the oxygen signal
convenience. The MAX30100 analog front end transmitted by IIC interface to the MCU for the further
contains two LEDs (Red (660nm) and infra-Red processing. What’s more, The SpO2 system in the MAX30100
(940nm)) and a photodiode. The MCU controls the is composed of ambient light cancellation (ALC) and
drive circuit to drive the red and infrared LEDs proprietary discrete time filter to reject 50Hz/60Hz
alternately according to the timing, and then the interference and low-frequency residual ambient noise [8].
photodiode will transform the reflective optical signal In the main control circuit section, we use the chip is
into electrical signals. After the analog signal Cortex-M3 STM32F103 series MCU, which has high
conditioning, a 16-bit internal successive- performance, low cost and low power consumption. In this
approximation analog-to-digital converter (ADC) system, the microprocessor main tasks are as follows: control
digitizes the analog signal. Finally, the digital signal
the LED and pulse length, process digital signals, calculate the system is powered on, the software will get the state by
value of blood oxygen saturation and pulse rate. Once the judging the flag, if the system detects no measurement, it will
MCU is working, the data of SpO2 and pulse rate can be sent go into ultra-low power sleep mode, in which almost no power
to the smart phone through Bluetooth 4.0 BLE interface. is consumed. Otherwise, the system will collect and process
B. Software the PPG signals and calculate the SpO2 and Pulse Rate values.
Finally, the system sends data to the smartphone APP through
Software design is the core of system, low power Bluetooth 4.0 BLE interface. The program flow diagram is
consumption is considered as an important factor. When the shown as Fig. 5:

6WDUW

6\VWHP LQLWLDOL]H

12
6WDWH JHW" /RZ SRZHU PRGH

<(6
0D[30100 LQLWLDOL]H

*HW 33* VLJQDO

Signal processing
12 12

*HW 3XOVH 5DWH" Algorithm *HW 6S2²"

<(6 <(6
%OXHWRRWK WR
VPDUWSKRQH

12
0HDVXUHPHQW (QG"

<(6
(QG

Fig.5 The software flow diagram of system

The pulse wave frequency is usually below 10Hz, which

mainly at 0.5Hz ~5 Hz [10-11]. There are usually two kinds of
C. Signal Processing digital signal processing methods are used: band-pass filter
The key to measuring the blood oxygen saturation is to denoising and wavelet denoising. However, the wavelet
obtain a smooth PPG signal and PPG signal feature point transform process is complex and is not conducive to MCU
extraction. Based on the optoelectronic pulse signal acquisition implementation. What's more, the selection of denoising
system, the reflected pulse signal has a small amplitude and is threshold based on wavelet transform is subjective, which will
disturbed by body movement and external environment in the cause a lot of uncertainties in practical applications. Therefore,
process of signal acquisition. It is very easy to introduce noise. the system adopts band-pass filtering method to eliminate
The noise mainly includes baseline drift caused by respiration noise interference. The filtering effect is shown in Fig.7.
and muscle jitter, 50Hz frequency interference, white noise
The signal after band-pass filtering
and motion artifact and so on[9]. The original PPG is shown in 0.02
Fig.6. So signal processing is a process that is critical to the
PPG feature points. 0.01
Voltage(V)

The original PPG signal 0

-1.6
Voltage(V)
 -0.01

250 300 350 400 450 500

Sample(s)
Fig.7 The signal after band-pass filtering
-1.65
D. Peak Detection
0 100 200 300 400 500 600
Sample(s)
According to the oxygen saturation estimation formula, the
Fig. 6 The original PPG signal
SpO2 level mainly depends on the R value, which is computed
from the DC component (non-pulsatile) and AC component increasing is obviously, while the other area is relatively
(pulsatile) of two different PPG signals. In experiment, the smooth. So first, we need to perform the first-order differential
pulse wave DC component is obtained by averaging the signal processing of the oxygen signal, and set a suitable threshold
and the pulse wave AC component is equal to the wave peak according to the maximum and minimum of the post-
value minus the trough value. However, the phenomenon that differential signal. Second, the maximum value of the
the value of diastolic peak is over the value of systolic peak, differential signal is obtained by the threshold method, and
which caused a lot of errors in the main peak point extraction then range to find the right crest point of the PPG signal.
process. To solve this problem, this system adopts the Finally, the trough is position that the minimum point between
algorithm is Difference Threshold Method (DTM). In a pulse two crests.
wave cycle, the difference value of the main peak is Through experimental analysis, the DTM can find the
significantly larger than the other areas [12]. characteristic points of the pulse wave quickly and accurately
As shown in Figure 8, the thin line is a pulse wave signal, without affect other features of the original signal. Moreover,
the thick line is the pulse wave signal of differenced, it find the algorithm is simple to operate and easy to implement, and
that the differenced value of the main wave in the process of the program code is efficient.
Crests and Troughs
0.03 the PPG of filtered the PPG signal of differenced
0.02
Voltage(V)

0.01
0
-0.01
200 250 300 350 400
Sample(s)

Fig. 8 Crest and trough detection

Tab.2 Comparison of measurement result of Pulse Rate

E. System Test Num. Standard Measurement Relative error(%)

After the system is implemented, we compare the system 1 70 70 0
test results with MEC-2000 test results. The prototype device is
shown in Figure 10. MEC-2000 portable multi parameter 2 64 63 1.562
monitor of MINDRAY company for the medical department of
the ECG, blood pressure, oxygen saturation, temperature and 3 69 67 2.898
breathing and other vital signs were detected, which is one of
4 73 72 1.369
the multi parameter monitor is widely used. In this paper, we
use MEC-2000 measurements as standard values, and this 5 77 75 2.597
system measures the value as measurements. During the
measurement, the tester is required to be in a relatively static 6 66 65 1.515
state, we selected 6 volunteers for data collection, data records
are as shown in Tab.1 and Tab.2.
From the system prototype test results can be seen, the Bluetooth MCU
relative error between the measured value and the standard
value is within 3%. It can be proved that the reflective blood MAX30100
oxygen saturation detection system has feasibility.
By using the Bluetooth technology, the SpO2 value, Pulse
Rate and pulse wave can be real-time displayed on smart
phone. The running interface of APP is shown as in Fig.11.
7HVW 6\VWHP
Tab.1 Comparison of measurement result of SpO2

Num. Standard Measurement Relative Error(%) Fig.10: The experimental device

1 98 96 2.040

2 97 95 2.061

3 97 96 1.030
4 96 96 0

5 98 97 1.020

6 95 94 1.052

Fig.11: The running interface of APP

 IV. CONCLUSION
In this paper, based on the principle of oxygen saturation
measurement, taking into account the characteristics of
portable, low power consumption, we provide a reflection of
oxygen saturation detection system design based on the
MAX30100. The scheme includes hardware and software
structure design, signal processing method and system
effectiveness. In addition, we compared the detection system
with the market standard oximeter MEC-2000, the results
show that oxygen saturation and pulse rate have a certain
reliability. The system’s adopting of integrated chip brings the
advantages of small volume, low cost and easy portability.
Therefore, this design provides the basis for the study of the
reflective oxygen saturation detection system.

V. ACKNOWLEDGMENT
This work was supported by the National High Technology
Research and Development Program (SS2015AA020109) and
Guangdong Applied Research and Development Program
(2015B010129012), and the Engineering Laboratory of
Shenzhen Municipal Development and Reform Commission.

VI. REFERENCES
[1] Rotariu, C. and Manta, V.. “Wireless system for remote monitoring of
oxygen saturation and heart rate,” computer Science and Information
Systems(FedCSIS), 2012 Federated Conference on vol. 9, no.12, pp. 193-
196, 2012.
[2] Z. Zhang. “Photoplethysmography-Based Heart Rate Monitoring in
Physical Activities via Joint Sparse Spectrum Reconstruction,” IEEE
Trans Biomed Eng, vol.62, no. 8, pp. 1902-1910, 2015.
[3] M. Nitzan, S. Noach, E. Tobal. “Calibration-Free Pulse Oximetry Based
on Two Wavelengths in the Infrared — A Preliminary Study,” sensors,
vol. 14, no. 4, pp. 7420–7434, 2014.
[4] Sahoo, Ashima, and G. S. Rath. “Real Time Estimation of Various
Physiological Parameters by Analysis of PPG Signal,” International
Journal of Computer Applications, vol.57, no.15, pp. 11-15, 2012.
[5] Mannheimer, P. D.. “Wavelength selection for low-saturation pulse
oximetry,” IEEE transactions on bio-medical engineering, vol. 44, no. 3,
pp.148 – 158, 1997.
[6] Teja R. B. “Calculation of Blood Pulse Transit Time from PPG,” A
Bachelor of Technology in Biomedical Engineering Thesis Paper, 2012.
[7] Stuban, Norbert and Niwayama, Masatsugu, “Optimal filter bandwidth
for pulse oximetry,” Review of Scientific Instruments, vol. 83, no. 10, pp.
104708-104708, 2012.
[8] Maxim Integrated. Max30100. Pluse Oximeter and Heart-Rate Sensor IC
for Wearable Health. Maxim Integrated Products, inc.2014.
[9] Kan L, Jianqing L, Jianfeng W. “A cascade filter for pulse wave baseline
drift elimination,” Image and Signal Processing (CISP), 2012 5th
International Congress on. IEEE, pp. 1495-1499, 2012.
[10] J. M. Cho, Y. K. Sung, K. W. Shin. “A preliminary study on
photoplethysmogram(PPG) signal analysis for reduction of motion
artifact in frequency domain,” Biomedical Engineering and Sciences
(IECBES), 2012 IEEE EMBS Conference. pp. 28–33, 2012.
[11] Shi Longfei, Zhao Ke, Li Ye. “Design of Noninvasive Blood Oxygen Saturation
Measurement System Based on AFE4400,” Integration Technology, no. 02, pp. 75-
85, 2015.
[12] Zhao Aihua, Wang Ping, Duan Yongxin. “Numerical detection algorithm of pulse
signal based on dynamic difference threshold,” Journal of Jilin University
(Engineering Science), no. 03, pp. 847-853, 2014 .