Hindawi

Computational and Mathematical Methods in Medicine

Volume 2021, Article ID 7152576, 16 pages
https://doi.org/10.1155/2021/7152576

Research Article
Development of an IoT-Based Sleep Apnea Monitoring System for
Healthcare Applications

Abdur Rab Dhruba ,1 Kazi Nabiul Alam ,1 Md Shakib Khan ,1 Sami Bourouis ,2
and Mohammad Monirujjaman Khan 1
1
Department of Electrical and Computer Engineering, North South University, Bashundhara, Dhaka 1229, Bangladesh
2
Department of Information Technology, College of Computers and Information Technology, Taif University, P.O. Box 11099,
Taif 21944, Saudi Arabia

Correspondence should be addressed to Mohammad Monirujjaman Khan; [email protected]

Received 23 September 2021; Revised 9 October 2021; Accepted 16 October 2021; Published 3 November 2021

Academic Editor: Osamah Ibrahim Khalaf

Copyright © 2021 Abdur Rab Dhruba et al. This is an open access article distributed under the Creative Commons Attribution
License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is
properly cited.

Sleep is an essential and vital element of a person’s life and health that helps to refresh and recharge the mind and body of a
person. The quality of sleep is very important in every person’s lifestyle, removing various diseases. Bad sleep is a big problem
for a lot of people for a very long time. People suffering from various diseases are dealing with various sleeping disorders,
commonly known as sleep apnea. A lot of people die during sleep because of uneven body changes in the body during sleep.
On that note, a system to monitor sleep is very important. Most of the previous systems to monitor sleeping problems cannot
deal with the real time sleeping problem, generating data after a certain period of sleep. Real-time monitoring of sleep is the
key to detecting sleep apnea. To solve this problem, an Internet of Things- (IoT-) based real-time sleep apnea monitoring
system has been developed. It will allow the user to measure different indexes of sleep and will notify them through a mobile
application when anything odd occurs. The system contains various sensors to measure the electrocardiogram (ECG), heart
rate, pulse rate, skin response, and SpO2 of any person during the entire sleeping period. This research is very useful as it can
measure the indexes of sleep without disturbing the person and can also show it in the mobile application simultaneously with
the help of a Bluetooth module. The system has been developed in such a way that it can be used by every kind of person.
Multiple analog sensors are used with the Arduino UNO to measure different parameters of the sleep factor. The system was
examined and tested on different people’s bodies. To analyze and detect sleep apnea in real-time, the system monitors several
people during the sleeping period. The results are displayed on the monitor of the Arduino boards and in the mobile
application. The analysis of the achieved data can detect sleep apnea in some of the people that the system monitored, and it
can also display the reason why sleep apnea happens. This research also analyzes the people who are not in the danger of
sleeping problems by the achieved data. This paper will help everyone learn about sleep apnea and will help people detect it
and take the necessary steps to prevent it.

1. Introduction tive sleep apnea (OSA) is basically upper airway congestion.

It occurs due to relaxation of our throat muscles and lack of
Sleep apnea is a disorder which is parallelly connected to the oxygen passing through our nasal-throat passage that causes
human respiratory system and our brain. It indicates a prob- irritation in breathing and sometimes results in serious com-
lem with breathing and respiration obstacles during sleep. plications. Central sleep apnea is less popular but more
There are two kinds of sleep apnea: obstructive sleep apnea alarming than OSA. The human brain is the key maintenance
(OSA) and central sleep apnea (CSA). Sometimes they both center of the whole body, as every organ and system runs on
take place at the same time. That can be said as complex sleep its directed signal or instruction, and its central sleep apnea
apnea, but it is too rare to be called an apnea type. Obstruc- comes into major exposure. Central sleep apnea occurs when
2 Computational and Mathematical Methods in Medicine

the brain fails to send the necessary instructions or signals to based on sleep apnea. In [7], wearable e-textile sensors used
the system which controls our breathing and respiration. an IoT approach to collect real-time data on sleeping habits
Here, neurons fail to transmit signals to the breathing mus- and respiratory rate. The same concept is used in [8] as they
cle, which pauses the breath for a good amount of time, use signals of breathing collected from the sensor-based
maybe near 10 seconds [1]. And in rare cases, they can both mattress, which are generally used in operation theatres. A
take place, which is more than an emergency medical treat- polyvinylidene fluoride (PVDF) sensor is used with a poly-
ment issue, and alarming also. somnography (PSG) device to measure time-to-time AHI
Medical and clinical researches are going on regarding data [9]. Researchers used the “peripheral arterial tonome-
this issue, and some alarming scenarios are keeping us alert. try” (PAT) method to detect sleep apnea perfectly. They
Sleep apnea is responsible for a wide range of physical com- developed the WatchPAT device that records the pulse wave
plications and diseases, including strokes, hypertension, car- on a finger and derives sleep and sleep apnea features [10].
diac abnormalities, and depression [2]. According to In [11], a 3D camera was being used to check the movement
research, around 3 to 7% of men and 2% to 5% of women of abdominal muscles which they compared with the OSA
are suffering from sleep apnea. That sums up approximately detection process. The sound detection method used in
more than 100 million people in the world, including adoles- [12] is one that allows for precise measurement of sounds
cents. Interestingly, 80% of apnea cases remain undiagnosed. with very high frequencies. Besides, so much research has
Sleep apnea (OSA) affects approximately 1 to 4% of children taken place with the analysis of real-time ECG monitoring
aged 2 to 8, with 20% of them snoring. It varies in different and prereserved ECG data. Machine learning techniques
parameters, like who is affected, who is not [3, 4]. Some were implemented. In [13], the authors described the devel-
complications or risk factors related to this problem identi- opment of an android-based pulse monitoring system. To
fying process are excessive weight or obesity, being men measure the human heart rate, the author of this study used
(2-3 times higher risk), alcohol consumption, smoking, fam- only one sensor. In [14], an IoT device was built that extracts
ily history, neck circumference, nasal congestion, and medi- ECG data from the user and evaluates the result from the
cal conditions like high blood pressure, type-2 diabetes, and classification model- (SVM-) based system, which they
lung or other respiratory diseases. Patients with diseases like named the “Apnea MedAssist Service”. In [15–17],
Parkinson’s and after strokes are at risk of suffering from researchers implemented machine learning and deep learn-
central sleep apnea. Sleep apnea can cause a good number ing techniques to detect apnea from medical data.
of health complications and death in the very worst case. From earlier research, it is observable that researchers
Loud snoring, fatigue, drowsiness, weakness in the body, measured real-time parameters of sleep apnea, which they
and lack of concentration are the common symptoms of analyzed afterwards. But there is less use of parameters of
sleep apnea. High blood pressure caused by a lack of sleep sleep apnea like SpO2, heart rate, AHI, sleep time, or such
and a low oxygen level may also increase the risk of a heart important parameters being measured at a time. In particu-
attack. Apnea patients are three times more likely to have a lar, they measured data and analyzed it with advanced tools.
stroke. Lungs can be affected even more dangerously when Besides, in earlier research, these values were measured over-
the SpO2 level is reduced. Lung disability caused by a lack night, and then, they got the data, but in this system, the
of oxygen is a very common but very concerning complica- value will be measured in real time. If any complications
tion of sleep apnea. 43% of people suffering from mild sleep arise, someone concerned may be notified via IoT device.
apnea had “hypertension.” According to research, this apnea For example, if “oxygen saturation” falls, this is a severe
problem causes 15% of traffic accidents and costs approxi- medical emergency, and with a possible device, the atten-
mately 1000 lives in the United States. Over 38,000 people dant, doctor, or concerned personnel can get instant signals
in the USA die every year due to direct and indirect effects from this individual patient. Hence, this proposed system in
of sleep apnea, mostly affected by cardiovascular complexi- the research is novel work. Due to sleep apnea, people may
ties [5]. So, this is a matter of awareness as it is not being have a stroke, may have cardiac arrest, and may have heart
focused on yet or getting that much exposure, but we will failure or such complications. Real-time measurement of
try to identify it with ease with our research so that people these apnea parameters is a must to reduce such risks, as it
do not remain undiagnosed and can have a comfortable life may cause death. Another motivation is that regular apnea
with no life risk. screening devices are too costly to afford. They are mostly
The very basic parameters of health which are related to used in medical institutes for medical conditions or research
sleep apnea include AHI (apnea-hypopnea index), heart rate purposes, but a cost-effective device is needed so that the
(BPM), blood oxygen saturation (SpO2), body mass index general public can buy this and screen their apnea issues
(BMI), sleeping time, REM, blood glucose, age, and blood while staying at home, with no need to rely on costly clinical
cholesterol [6]. In different research, it has been seen that monitoring. Our research goal is to solve these issues, and
such parameters are covered on the basis of medical saving human lives is the ultimate goal of this study.
research. In most cases, they come up with a study of med- The objective of this study is to develop an IoT-based
ical data of a particular or group of patients, and they ana- sleep apnea monitoring system for health care applications.
lyze the results. Devices are mostly used: polysomnography An IoT-based sleep apnea monitoring system for healthcare
(PSG), electrocardiogram (ECG), and some clinical devices applications is developed in this study. In this research, the
like this. These devices are too costly to use for personal system will monitor the parameters of sleep apnea on a
use. Some advanced and fruitful research happened earlier real-time basis and will try to monitor the values of these
Computational and Mathematical Methods in Medicine 3

as efficiently as possible with diverse parameters. The main troller has the capacity to connect to other sensors or com-
parameters covered in this research are SpO2, heart rate, puters, allowing it to be used in many projects. It has 2 KB
ECG, skin response, and sound intensity to be measured at of SRAM (static random access memory) and 32 KB of flash
a time. An application is being made that will show the memory, 13 KB of which is utilized to store the set of
real-time data of these parameters so that patient’s attendant instructions in the form of code. It also includes a 1 KB
can get an instant and continuous update of patient’s health. EEPROM (electrically erasable programmable read-only
The rest of this paper is organized as this pattern: Section memory).
2 describes the methods and tools implemented in our This Arduino board has a total of 30 connections, with
research. All the modules (hardware and software) and their 14 digital pins and 6 analog pins for external connection.
workflows are being discussed here. Section 3 demonstrates The A0 to A5 analog pins are used to receive analog data
the results and outcomes we got from our system and their from external devices such as analog sensors. On the board,
broad qualitative analysis as well. In Section 4, the whole there are various digital and analog input and output pins
concept of this research is summarized, and we also dis- that operate at 5 V. These pins have conventional opera-
cussed how important and effective it is to monitor sleep tional current ratings of 20 to 40 milliamps. The DC power
apnea on a real-time basis, and we concluded the section jack may provide a voltage ranging from 7 V to 20 V, or
by discussing our ambitions and motivation for this work the USB connected to an external device can provide a 5 V
for the sake of lives. voltage. Data transmission is the key for IoT devices. To
receive, transmit data, and maintain serial communication,
2. Methods and Materials two pins called Pin 0 (Rx) and Pin 1 (Tx) work simulta-
neously. The Rx pin receives data, whereas the Tx pin trans-
2.1. Methodology Statements. This section discusses the mits data. Serial communication can also be done by other
methods, components and paths that are used to fulfill the input/output (I/O) pins of the board. The serial monitor of
goal. The aim of the system is to monitor the entire sleeping the Arduino IDE software is used to send or receive text data
period of a patient with sleeping disorders. This from the board. It is used to display the output data from the
microcontroller-based sleep apnea monitoring system for Arduino board.
sleeping disorder patients is combined with three different
layers. The main layer is a microcontroller unit which con- 2.3.2. Sensors. There are several sensors to monitor patient’s
nects the input layer and the output layer. In the input layer, condition during sleep. To monitor and analyze sleep apnea,
it is combined with four different sensors which will provide the system contains a few sensors that will fulfill the major
the analog signal to the Arduino UNO to measure the differ- goals. The heart rate module, SpO2, pulse sensor, ECG sen-
ent indexes of sleep condition. The output layer is combined sor, galvanic skin response (GSR) sensor, and sound sensor
with two parts, including the serial monitor of the Arduino to monitor the snoring sound of the patient during sleep
UNO and a mobile application to display the digital data have been used in the system. These six sensors will monitor
converted by the microcontroller. the patients during sleep and will provide analog data to the
Arduino UNO.
2.2. Outline of the System. A block diagram shows the full
system in Figure 1. The system consists of input, output,
(1) Heart Rate Pulse Sensor. Difficulties with breathing and
and a microcontroller board Arduino UNO shown in
sleep apnea problems are very much related to the heart rate
Figure 2. The Arduino board, which is also connected to
of the patient. The system contains a heart rate pulse sensor
the output layer, combined with the serial monitor of the
to monitor the heart rate of the patient throughout the sleep-
Arduino board, and a Massachusetts Institute of Technology
ing period. A high heart rate is one of the major risk factors
(MIT) App Inventor-based mobile application connected to
for sleep apnea. The normal heart rate of a healthy person is
a Bluetooth module show the converted digital data to the
in the range of 60-100 beats per minute (bpm). The heart
viewer.
rate varies from person to person. More physically active
Figure 1 shows the basic workflow of the system. The
people usually have a lower heart rate than less active people.
sensors provide the data to the Arduino UNO simulta-
A higher and unusual change in heart rate is a major sign of
neously, and the Arduino UNO passes the converted digital
sleeping disorders. Figure 3 shows the heart rate pulse
data to the Arduino IDE (Integrated Development Environ-
sensor.
ment)’s serial monitor and also to the mobile application
through the Bluetooth module at the same time.
The heart rate during sleep is usually lower than at other
2.3. Modules and Materials. The system is integrated with times of the day. The average heart rate of a person during
different kinds of components that are doing different tasks sleep is usually around 60-80 bpm, as shown in Figure 4,
in the system. Some are for the input, some are for the out- but a healthy and fit person has a lower heart rate in the
put, and some are used in the system to create a bridge range of 50-60 bpm during the sleeping period.
between the inputs and outputs. An unusual change in heart rate is one of the symptoms
of sleep apnea. A lower heart rate can occur due to a lack of
2.3.1. Arduino UNO. The Arduino UNO, shown in Figure 2, oxygen supply in the body. It could happen when the breath-
is system’s major component, and it is based on the VR ing of a person stopped due to any unusual reason. A higher
Microcontroller Atmega328. This programmable microcon- heart rate is also a result of sleep apnea. Because of
4 Computational and Mathematical Methods in Medicine

Max 30102
Viewing outputs Arduino monitor

ECG sensor
Arduino UNO Bluetooth
module
GSR sensor
Viewing outputs

Mobile
Sound sensor application

Figure 1: Full system block diagram by which the system works successfully.

DC power jack
USB port

Reset button

Figure 2: Arduino UNO Pinout [18].

can provide the state of the heart during the sleeping period,
whether there is any kind of enlargement of the heart occur-
ring due to hypertension, and can also detect myocardial
infarction. The ECG varies person to person for several rea-
sons. A physically active person has a more stable ECG than
an inactive person’s ECG. The resting ECG is different from
the stress or exercise ECG. Figure 5 shows the AD8232 ECG
sensor used in this research. Figure 6 given below will show
the ECG values during sleep.

Figure 3: Heart rate pulse sensor. The system will contain an AD8232 ECG analog sensor
to monitor the ECG of a person shown in Figure 5. That sen-
hypertension and stress, a higher heart rate can occur. And it sor will be connected to the body and pass an analog signal
usually hampers the usual sleep of a person. It is one of the to the Arduino to convert the signal into a digital signal. This
main reasons for heart failure during sleep. sensor is connected with the Arduino via five pins. Among
The system will include a heart rate sensor to continu- them, two are for powering up the sensor. The input voltage
ously monitor person’s heart status. This is an analog sensor, of the sensor is 3.3 V. Other pins are the data pin, which is
and the sensing element of the sensor will be attached to per- connected to the analog pin of the Arduino, and others are
son’s body. This will send the analog signal to the Arduino the lead pins LO+ and LO-. This lead pin analyses the elec-
through the analog pin. The heart pulse sensor is depicted trode which is connected to the body and gives the data to
in Figure 3. The data pin of the sensor is connected with the Arduino. These ECGs can be excessively noisy at times.
one of the analog pins of the Arduino for data transmission.
This sensor operates on a voltage of 5 V and 4 mA current. (3) Max 30102 Finger Oximeter Heart Rate Module SpO2.
The amount of oxygen circulating in the blood is referred
(2) AD8232 ECG Sensor. An electrocardiogram, or ECG, is to as the blood oxygen level. Oxygen distribution throughout
very important in monitoring sleep apnea. An ECG provides the body is an indicator of detecting a healthy body and an
information about the heart rate and the rhythm of the unhealthy body. The blood oxygen level or oxygen satura-
heart. It also shows the unusual change in the heart rate. It tion level indicates how equally oxygen is distributed in the
Computational and Mathematical Methods in Medicine 5

Subject A : Healthy
100 Mean = 64.98

Heart rate (bpm)

90 SD = 4.79
80
70
60
50
0 5 10 15

Time (min)
Subject B : Obstructive sleep apnea
100
Heart rate (bpm)

Mean = 64.95
90 SD = 4.68
80
70
60
50
0 5 10 15

Time (min)

Figure 4: Heart rate during sleep [19].

saturation values fall from these threshold values. Figure 7

shows the oxygen saturation during a sleep apnea situation.

To measure the blood oxygen level, the system contains a

sensor-based pulse oximeter. The MAX30102 sensor will be
used in the system to monitor the oxygen saturation in blood
shown in Figure 8. This sensor combines two light-emitting
diodes (LED), a photo detector, optimized optics, and low-
noise analog signal processing to detect pulse oximetry.
Figure 5: AD8232 ECG sensor.
Figure 9 given below shows the system block diagram of
MAX30102.
Waking The main mechanism behind measuring the SpO2 value by
ECG
this sensor is to detect the ratio of blood-carrying hemoglobin
and blood-not-carrying hemoglobin. This sensor can measure
EMG/ECG the SpO2 value from the fingertip of a person. It operates on a
voltage ranging from 3.3 V to 5 V and senses the data from per-
UNIT son’s body and passes it to the Arduino to convert the data.
Sleep
(4) Galvanic Skin Response (GSR) Sensor. The galvanic skin
ECG response or GSR, also known as emotional arousal, refers
to variations in sweat gland activity that reflect the degree
EMG/ECG of our emotional state [22]. It indicates the electrodermal
activity of a person’s body. It is the result of sweat glands
UNIT in the skin being activated by the autonomic nervous system.
Figure 6: ECG curves. The skin response is different at different times of the day. It
also changes with the state of mind of a person. A stressful
person tends to have a lower skin response than a joyful per-
body from the lungs to the cells. Problems in the lungs are son. The GSR value also tends to be lower during sleep.
one of the major reasons for low oxygen saturation. Breath- Sweating hands is also a factor in skin response. Excessive
ing problems can cause a low supply of oxygen to the lungs sweating is one of the main results of hypertension. It
and the body. During OSA, the breathing rate usually falls. reflects the lower value of skin response.
This causes a lower supply of oxygen to the body. That
change affects the rate of oxygen saturation of blood. A The value of GSR is measured by the conductance of the
healthy person’s oxygen saturation, which is also referred skin. The equation to measure skin response is
to as SpO2, is 95% or higher. A sleep apnea patient usually
has a lower SpO2 value which is usually around 90%. Lower  
1
SpO2 can cause hypoxemia. It also indicates chronic lung Conductance = : ð1Þ
disease. Patients need external oxygen supply if the oxygen Resistance
6 Computational and Mathematical Methods in Medicine

tant. An unusual change in snoring replicates the intensity of

the sleep apnea problem. Figure 12 shows the sound sensor
module.

To measure the intensity of the snoring, the system will

contain an analog sound sensor that will measure the snor-
ing sound and determine the snoring during the sleeping
period. The sound sensor is a compact board with a micro-
phone and some processing circuitry for translating sound
waves into electrical signals. In the system, the sensor is con-
nected to the Arduino via the analog pin of the Arduino.
Arduino gets the analog data from the sensor and translates
it into digital for the mobile application.

Figure 7: Max30102 SpO2 sensor. 2.3.3. Bluetooth Module. Serial communication is the key to
this IoT-based paper. For that, the system contains a Blue-
Oxygen saturation during an apnea event tooth module HC-05. This Bluetooth module is the gateway
96 between the Arduino Uno and android application. Oper-
Oxygen saturation (%)

95 ated in two moods, this Bluetooth module sends or receives

94 data to another device in one mode, and another mode is
93 working in AT command mode to set device’s settings as
92
default. After pairing with the Bluetooth device, the digital
91
90
data will be visible in the mobile application for the user.
89 Figure 13 shows the pin out of the HC-05 Bluetooth module.
0 10 20 30 40 50 60 The devices operate in a voltage range of 4 V to 6 V and a
current of 30 mA. The TX (transmit) and RX (receive) pins
Time (sec)
of the Bluetooth module operate serial communication. To
Oxygen level transmit serial data, the TX pin operates, and the RX pin
Threshold value works to receive data from the microcontroller. The TX
pin of the Arduino is connected to the RX pin of the Blue-
Figure 8: SpO2 level during sleep apnea [20]. tooth module, and the RX pin of the Arduino is connected
to the TX pin of the Bluetooth module.
In the system, after pairing with the Bluetooth module,
The value of skin response is measured in micro- the mobile application, digital data is transmitted to the
Siemens. The average value of the GSR of a normal person app from the Arduino.
is in the range of 250 μS to 450 μS. Sleeping GSR usually falls
as the period of sleep increases. But any unusual change in 2.3.4. Components and Cost. The system is quite cost-
skin response can cause major issues. A sudden drop in effective for detecting sleep apnea. Table 1 shows the list of
response is a symptom of sleep apnea as it can occur with components required to build the system and the cost of it.
sudden emotional state changes caused by stress or hyper- The only cost of the system is for the hardware components,
tension. Figure 10 shows the GSR sensor module, and especially the sensors. Some of the sensors are priced at
Figure 11 shows the changes of electrodermal activity during more than 850 Bangladesh taka (approximately 10 US dol-
sleep. lars). The total cost of the system is 4570 taka ($53.76).
To measure the galvanic skin response during sleep and Although the system is quite cost-effective, it is very useful
monitor sleep apnea, the system contains a wearable analog for monitoring sleep apnea. One patient at a time can utilize
sensor. The galvanic skin response sensor enables the system this device, and it can be used multiple times as per the
to monitor sweat gland activity, which is linked to emotional requirements.
arousal. This sensor is easily wearable on the finger to mea-
sure the GSR value. The sensor passes the value to the Ardu- 2.3.5. MIT App Inventor 2. This system contains a mobile
ino to convert it into digital data, and later, it passes to the application to show the real-time data transferred from the
mobile application. Arduino via Bluetooth. This application shows all of the
parameters related to sleep apnea continuously. In the sys-
(5) Sound Sensor. Snoring is one of the major symptoms of tem, the name of the application is “sleep apnea monitoring
obstructive sleep apnea or OSA. It indicates the problem device.” This mobile application was created using MIT App
with breathing. Snoring is a raspy or loud sound produced Inventor 2 [22]. It allows the user to create an application
by air passing through relaxed tissues in the throat, causing and also has the functionality to create a gateway between
the tissues to vibrate while you breathe. From a health stand- the hardware devices. This mobile application shows the dig-
point, OSA-related snoring is more concerning. To detect ital data converted by Arduino UNO. To do that, the Blue-
OSA, monitoring of snoring sound and frequency is impor- tooth module HC-05 is connected to the Arduino. After
Computational and Mathematical Methods in Medicine 7

Cover glass No ink

10
ADC

Control Signal
Hbo2 processing

Hb

0.1
RED IR

Figure 9: System block diagram of MAX30102 [21].

and the GSR sensor. Also, the data conversion and transmis-
sion parts are performed in this section, and the data passed
to the serial monitor, and mobile application happens in this
part. After pairing the Bluetooth module with the mobile,
the data is displayed on the serial monitor and also stored
for detailed analysis.

2.3.7. Diagrams Drawn Using Software Showing the Layout

of the System. The circuit diagram with all the sensors and
Arduino Uno is designed using circuito.io [24], an online
IoT design platform. This website allows you to design
Figure 10: GSR sensor module. mostly Arduino-based devices. The hardware part of the sys-
tem was created on the basis of this design. A circuit diagram
of the system is given in Figure 16.
pairing with the mobile and connecting to the application,
The hardware setup of the study was done by following
the values become visible in the mobile application. The lay-
the diagram shown in Figure 16. The sensors are connected
out of the application that we made in the MIT App Inven-
with the Arduino UNO in the way displayed in the diagram.
tor is shown in Figure 14.
The data pin of all of the sensors is connected to the analog
pin of the Arduino individually, as shown in the figure,
2.3.6. Full System Review and Working Process Flowchart. which is also connected to the positive and negative pins of
The system is combined with a few different steps. Every the breadboard to get power. The Bluetooth module is con-
step has a different task to do, and both make it useful for nected to two of the I/O pins of the Arduino to continue
the user. Our whole hardware setup is shown in Figure 15. serial communication.
The programmable Arduino UNO program combines those
different steps. A flowchart of the Arduino program is 3. Result and Analysis
shown in Figure 16. The Arduino program is combined with
different functions for the different sensors and Bluetooth In this investigation, it has been discovered and managed to
modules that connect the Arduino UNO with the mobile deconstruct a number of outcomes. In the next sections, the
application. results and observations will be discussed in more detail.
System’s work flow started from the Arduino IDE after
uploading the code to the Arduino UNO. The code contains 3.1. Patient List of Examination. To detect sleep apnea in
the setup of the sensors, the working process of the sensors, patients, the system monitored different people from differ-
and the Bluetooth module. The main function of the code is ent age groups and different health conditions. The system
controlled by a loop which allows the system to display the examined both male and females and got to see quite
data every second. The loop controls the organization of remarkable differences. The device attempted to examine a
the Max30102 module, the ECG module, the sound sensor, patient with various health issues because previous health
8 Computational and Mathematical Methods in Medicine

Electro dermal activity in SLEEP

Skin resistance

Falling asleep Sleep

1st cycle 2ndcycle

Awake 1st stage

REM
Relaxation
2nd stage
SWS

Active awake
GSR
GSR Time

Figure 11: ED activity during sleep [23].

Figure 12: Sound sensor.

Figure 13: Bluetooth module.

Table 1: List of components and cost.

Name of components Units Cost (BD taka)

Arduino Uno 1 400 Taka ($4.7)
Heart rate sensor 1 300 Taka ($3.53)
Figure 14: Sleep apnea monitoring app in the MIT App Inventor
AD8232 ECG sensor 1 950 Taka ($11.18) website during development.
MAX30102 1 850 Taka ($10)
Sound sensor 1 150 Taka ($1.76)
GSR sensor 1 1200 Taka ($14.12) reports and disease history are one of the factors contribut-
HC-05 Bluetooth module 1 300 Taka ($3.53) ing to sleep apnea. The focus was on the body condition of
Breadboard 1 90 Taka ($1.05) the person before monitoring sleep apnea. The IoT device
Jumper wires 1 box 130 Taka ($1.53) monitors the patients and provides us the results against
all of the parameters. The list of persons examined by the
Others — 200 Taka ($2.35)
system is given below in Table 2 with their age group and
Total = 4570 Taka ($53.76)
gender.
Computational and Mathematical Methods in Medicine 9

Arduino UNO pin setup

Start
Blutooth module pin setup

Loop

Read ‘Analog’ data from Read ‘Analog’ data from

Max30102 sensor (Heart AD8232 sensor (ECG/EKG)
rate and SpO2) and convert and convertthem into
them into ‘Digital” data output ‘Digital” data output

Read ‘Analog’ data from

Read ‘Analog’ data from
Sound sensor (sound
GSR sensor (Skin response or body
intensity upon ‘snoring”)
resistance) and convert them
andconvert them into
into‘Digital” data output
‘Digital” data output

Recive and merge Bluetooth Showing ‘Real-time’

data from alla sensors on data in mobile App

Storing data Showing data in

for further arduino monitor
analysisi

Figure 15: Whole system working flowchart.

Figure 16: Circuit diagram of the full system.

10 Computational and Mathematical Methods in Medicine

Table 2: Lists of people that has been examined.

List of person Age group Gender Body status

Person 1 5-17 Male Normal
Person 2 18-35 Male Overweight, excessive sweating problem
Person 3 18-35 Male Physically fit (athlete)
Person 4 36-50 Female Major heart issue, obese
Person 5 50+ Male Lung issue

Figure 17: Prototype of the system.

3.2. Design Prototype and View of Real-Time Data. After

connecting all of the sensors to the Arduino and uploading
the code to the board, the values of all the parameters appear
on the screen and in the mobile application. The system
works successfully to monitor sleep apnea. Figure 17 shows Figure 18: Sleep apnea monitoring on a patient.
the hardware part of the prototype of the system.
The prototype system is integrated with the Arduino,
sensors, and the Bluetooth module. Five different sensors
are integrated in the system. Those sensors are marked in
Figure 17. The system includes a GSR sensor to measure
the skin response, a pulse sensor to measure the heart rate,
a Max30102 sensor to measure oxygen saturation (SpO2),
an ECG sensor to monitor the electrocardiograms of the
heart, and a sound sensor to measure the snoring intensity
integrated into the system. The HC-05 Bluetooth module is
in the system to perform serial communication between
the hardware part and the mobile application.
Figure 18 shows the prototype of the system during
monitoring of a patient. The prototype of the system is an
easily wearable device. It does not hamper the sleep of the
patient. The application shown in Figure 19 shows the
real-time results of oxygen saturation (SpO2), heart rate, gal-
vanic skin response, and sound intensity level. The mobile
application can show the ECG graphs of the patients. But
the serial plotter of the Arduino gives better ECG results.

3.3. Data Sheet and Analytical Charts for Different

Parameters. In the experiment, the system monitors person’s
sleeping period from some different perspectives and tries to
see how body’s condition changes during the sleeping Figure 19: Mobile application showing real-time result.
period. For that, it monitored the patient for six straight
Computational and Mathematical Methods in Medicine 11

hours and analyzed how different parameters changed every oxygen saturation during the third and fourth hours of sleep.
hour. That is one of the symptoms of breathing problems. A low
rate of breathing declines the rate of oxygen supply to the
3.3.1. Heart Rate Analysis. Table 3 shows the results of the body, and it decreases the level of oxygen saturation. That
heart rate for the five people. It shows how the heart rates low SpO2 value is an indication of obstructive sleep apnea.
of different people change over hours. The values indicate The average value of oxygen saturation for different peo-
how the heart rate of a person changes from the start of sleep ple shown in Table 4 shows some variations. Although the
to the end of sleep. In the experiment, we measured the value variation is not significant, for person 5, the value is a little
of heart rate every second throughout the period and lower, and for person 3, the value is higher than the other.
checked the highest, lowest, and value that clicked most of Person 5 has a mean SpO2 of 94.55, while person 3 has a
the time. After checking the values, we took the decision to mean SpO2 of 97.81.
calculate the average heart rate in each hour. The average The line chart for SpO2 in Figure 21 shows how the level
value reflects the heart rate that the person had at their max- of oxygen saturation changes over time more precisely.
imum time in that specific hour. In the experiment for every The line chart for the five people in Figure 21 shows the
person, the values are in some specific pattern, and in all change in oxygen saturation during different hours of the
cases, they follow the body condition of the person. Person sleeping period. It reflects how oxygen distribution happens
1, who does not have any major health issues, has a very bal- in the bodies of those five people over the hours.
anced heart rate throughout the sleeping period. The results
do not provide any indication of sleep apnea. The same goes 3.3.3. Galvanic Skin Response (GSR) Analysis. Galvanic skin
for person 3, who is an athlete and has very good physical response (GSR) values for the five people are shown in
condition. His heart rate is in the range of 67-76 beats per Table 5. It shows every person’s electrodermal activity
minute during the period. That value is quite standard for throughout the sleeping period. The value indicates how a
a healthy person. But, for other people, the results do show person’s state of mind changes over time during sleep
some of the major hints about sleep apnea. Person 2 has because of stress or hypertension. During the experiment,
an uneven change in heart rate. This person has a heart rate the system monitors person’s electrodermal activity via the
of around 90 bpm. In the third hour of sleep, his heart rate sensor and shows the response of the skin every second.
falls. One of the most serious consequences of breathing dif- The response of the skin changes gradually for every person.
ficulties is a low heart rate, also known as Bradycardia. In For person 1, the response of the skin is quite balanced
that case, that low heart rate indicates obstructive sleep throughout the sleeping time, and it does not show any hint
apnea in that patient. On the other hand, people 4 and 5 of sleep apnea, and also for person 3, as his skin response is
have a higher heart rate. According to the history, person 4 highest among the others. But for person 2, person 4, and
has a major heart problem, which suggests that a higher person 5, the result shows an uneven change in skin
heart rate could be the cause of both the heart problem response. For person 2, the value of GSR goes down in the
and obesity. 3rd hour and also in the last two hours. It shows the chance
The average value of BPM of test subjects varied from of stress and hypertension problems during sleep. Person 4
person to person for different ages, physical conditions, has the lowest GSR value among all of the people the system
and previous health issues. There is quite a massive variation monitored. In the first two hours, the values of skin response
in the BPM of different persons. From Table 3, it is noted do not show any change, which can lead to a decision about
that the mean value of the heart rate for person one is sleep apnea. But from the 3rd hour to the end of the sleep,
74.47, while for person 4, it is noticed to be 91.39. the GSR value reflects the change of very low dermal activity,
For comparison of heart rate among the people, which can be the reason for sleep apnea. Person 5 also has
Figure 20 shows how heart rate changes because of the phys- quite a low skin response, but the value throughout the time
ical condition of the people. does not change significantly, and in that case, there is no
The line chart of heart rate in Figure 20 shows the aver- evidence for sleep apnea for person 5.
age heart rate of a person changing over the course of an The mean of galvanic skin response for five people in
hour. In the line chart, it is visible that most of the people Table 5 shows significant variations in the skin response
have an uneven change in heart rate. The differences over the hour. The lowest value (162.5) of GSR is noticed
between the lines of one person and another person indicate for person 2, whereas the highest value (225.54) is noticed
the difference in their age group and body condition. for person 1. The line chart shows the change in galvanic
skin response for the five people over the hour in
3.3.2. SpO2 Analysis. Table 4 shows the oxygen saturation Figure 22 more precisely.
for the five people. It shows the change of oxygen saturation
over hours. In the experiment, the system calculates the 3.3.4. Sound Intensity Analysis. Snoring, which is a symptom
SpO2 value every second with the sensor-based pulse oxim- of obstructive sleep apnea, was detected in three of the
eter. The SpO2 value always stayed within a particular range people the system monitored. The snoring sound followed
for all the people. For that reason, we took the average SpO2 a pattern for each person. In the experiment to detect
value for every hour of the sleeping period. All of the people snoring, the sensor collects data every second. The result
have quite good oxygen saturation levels throughout the shows that, for a specific person, the sound follows a sim-
monitoring period, except for person 5, who has very low ilar rhythm throughout the sleeping time. The graph in
12 Computational and Mathematical Methods in Medicine

Table 3: Result of BPM.

Hour Person 1 Person 2 Person 3 Person 4 Person 5

1st hour 73.01 88.67 69.37 100.24 74.32
2nd hour 76.39 88.63 68.17 98.48 82.67
3rd hour 67.33 62.73 67.29 97.12 87.23
4th hour 72.82 83.21 75.87 79.54 89.56
5th hour 78.38 91.04 67.5 82.96 76.98
6th hour 78.87 85.34 73.68 90.02 73.57
Mean BPM 74.47 83.27 70.31 91.39 80.72

110
110
100.24 98.48
97.12 100
100 90.02
Beats per minute (BPM)

Beats per minute (BPM)

89.56 91.04
88.67 88.63
90 80
82.67 83.21 82.96
79.54 85.35
80 74.32 76.39 76.98 78.87
7587. 78.38 70
73.01 67.29 73.57
68.17 73.68
70 69.37 67.33
72.82
97.5 60
62.73
60
50
50
40
40
1st 2nd 3rd 4th 5th 6th

Hours
Person 1 Person 4
Person 2 Person 5
Person 3

Figure 20: Line chart for heart rate.

Table 4: Result of SpO2.

Hour Person 1 Person 2 Person 3 Person 4 Person 5

1st hour 95.56 95.90 97.12 95.71 95.87
2nd hour 95.97 95.79 98.11 96.45 94.15
3rd hour 95.68 95.96 97.70 96.73 93.78
4th hour 95.83 96.12 97.22 96.46 93.63
5th hour 96.04 96.00 98.37 97.49 94.12
6th hour 95.81 95.94 98.36 96.07 95.73
Mean SpO2 95.82 95.95 97.81 96.49 94.55

Figures 23(a)–23(e) depicts the snoring rhythms of the normal rate as the breathing cycle for person 5 took 4-6 sec-
people studied. onds. Both people faced problems with breathing during
Person 3 has a very balanced rhythm in the breathing sleep, which is a sign of sleep apnea.
cycles. There is no evidence of snoring visible in
Figure 23(c). The sound level stayed in the range of 35 dB- 3.3.5. ECG Analysis. Figure 24 depicts the heart rhythm or
40 dB, which is usually for room sound. But for person 4 ECG graph for test person 4. Person 1, person 2, person 3,
and person 5, there is a big problem with snoring, as it is vis- and person 5 have a very balanced ECG result. But the
ible in Figures 23(d) and 23(e). For person 4, the figure ECG graphs of person 4 show some uneven changes in the
shows that the snoring sound reached almost 60 dB, and rhythm of the heart.
every snoring cycle took 3-5 seconds. That is big evidence In Figure 24, there are six waves that are the top picks of
of problems with breathing and also a sign of obstructive the graph. Those spikes are also known as R-wave. It shows
sleep apnea. The same goes for person 5, as snoring intensity how time elapsed between endocardium to epicardium. It is
touches the 60 dB mark. The breathing rate is lower than the also helpful for some of the major heart issues. Loss of R-
Computational and Mathematical Methods in Medicine 13

100
100
83.21
98.37 98.36
98.11
97.7 80
98 88.63
97.22 97.49
97.12

Sp02 (%)
96.73

Sp02 (%)
96.45 96.45 96.67
95.87
99.71
95.96 96.12 95.73 96
96 95.97 95.94
95.68 95.96 95.81
95.56 95.83

94
94 94.15 93.78 94.12
93.63

40
92
1st 2nd 3rd 4th 5th 6th

Hours

Person 1 Person 4
Person 2 Person 5
Person 3

Figure 21: Line chart for oxygen saturation (SpO2).

Table 5: Result of GSR.

Hour Person 1 Person 2 Person 3 Person 4 Person 5

1st hour 231.03 181.81 348.2 188.88 190.54
2nd hour 235.47 254.87 269.35 150.03 216.47
3rd hour 264.67 134.34 265.41 74.22 142.79
4th hour 221.74 154.95 217.13 123.25 135.38
5th hour 193.13 123.82 253.77 119.51 147.90
6th hour 207.18 125.21 298.82 118.08 164.27
Mean GSR 225.54 162.5 275.45 128.95 166.23

348.2 350
350 330
330 310
310 290.82 290
69.37 269.35
290 270
Micro siemens (us)

Micro siemens (us)

265.41
270 83.21 250
254.87 264.62 253.37
250 230
231.03 235.47
230 276.39 217.13 210
210 190.54 207.18 190
190 181.81 164.27 170
170 154.95 147.8 150
150.03 142.79
150 134.34 130
130 123.25
121.95
129.2
129.08 110
110 90
90 74.22 70
70 50
50
1st 2nd 3rd 4th 5th 6th

Hours

Person 1 Person 4
Person 2 Person 5
Person 3

Figure 22: Line chart for galvanic skin response.

wave indicates myocardial infarction, which is visible in the toring sleep apnea with similar kinds of parameters shown in
first R-pick in Figure 24. The gain of R-wave indicates car- Table 6. It also shows the parameters that were not used in
diomyopathy, which is the 5th R-pick of the graph. other research.
Most of the previous research to monitor sleep apnea is
3.4. Comparative Analysis. The outcome of the system has mainly based on one parameter. But this study combined
been compared to some other research that has been moni- five different parameters of the sleeping index. In this study,
14 Computational and Mathematical Methods in Medicine

Sound intensity (dB) Sound intensity (dB)

75 75

60 60

45 45

30 30

15 15

0 0
2 4 6 8 10 12 14 16 18 Time (S) 2 4 6 8 10 12 14 16 18 Time (S)
(a) (b)
Sound intensity (dB) Sound intensity (dB)

75 75

60 60

45 45

30 30

15 15

0 0
2 4 6 8 10 12 14 16 18 Time (S) 2 4 6 8 10 12 14 16 18 Time (S)

(c) (d)
Sound intensity (dB)

75

60

45

30

15

0
2 4 6 8 10 12 14 16 18 Time (S)
(e)

Figure 23: Snoring cycle of 5 persons; (a), (b), (c), (d), and (e) show the sound intensity of person 1, person 2, person 3, person 4, and
person 5, respectively.

450.0

300.0

150.0
2.637 2.737

Figure 24: ECG graph of test person 4.

Computational and Mathematical Methods in Medicine 15

Table 6: Comparison table with previous studies.

Reference Main parameters Result (final or mean)

In the paper (91:3 ± 2:0) %
Oxygen saturation (SpO2)
[25] The oxygen saturation for sleep apnea patients is in this range
In the paper (70:9 ± 2:2) bpm
Heart rate
[6] The study detected sleep apnea in this range of heart rate.
In the paper
Polysomonography 91.8% to 96.5% accuracy on adaptive apnea index
[26]
In the paper
Snoring sound intensity 45 dB to 58 dB and mean value is approximately 52 dB
[27]
In the paper A novel approach has been proposed to measure respiratory movement using
3D camera
[11] a 3D camera to detect apnea.
SpO2: 94.55 to 97.81
BPM: 74.47 to 91.39
SpO2, heart rate, ECG, skin response,
Our system GSR: 128.95 to 275.45
snoring sound intensity
Sound: 35 dB to 60 dB
ECG: generalized

the decision was made after analyzing all the five parameters ing systems. Some other sensors could be added to the sys-
combined as one factor is dependent on another factor of tem to make it more applicable. An LM35 temperature
sleep. Another important aspect of this study is no other sensor could be added to the system. A DHT11 humidity
research focused on the galvanic skin response to detect sensor could also be added to this system to make this sys-
sleep apnea. However, this study establishes a link between tem more applicable, though those factors have a minor con-
skin response and sleep apnea monitoring. That opens an nection with sleep apnea.
important area for analyzing sleep apnea more precisely. In most research, ECG data and other data are used to
measure or monitor sleep apnea. Most were under medical
4. Conclusion observation in a clinical environment. But our device will
help patients to get even better output while staying at home
This research shows how IoT devices can monitor sleep and with ease of use. Real-time data will give instant updates
apnea. To implement the system, we used a basic microcon- about apnea indexes so that patients can take precautionary
troller and some of the major health-related sensors. The steps. And for senior citizens and the physically/mentally
mobile application has been created with a very simple app disabled, our device is an added benefit because it sends
developing web application. After monitoring five people, notification data to patient’s concerned person so that the
the system gives quite satisfactory results for making deci- person can take care of any difficulties that may occur at
sions about sleep apnea. From the given results, it is clear any time. We will further try to build our own dataset with
that two people do not have any symptoms of any kind of this real-time data and will analyze it through multiple
sleep apnea. One person who is in the age range of 36-50 machine learning models to analyze the whole system and
has major issues with sleep conditions. The system success- how our system can work with better efficiency. To have a
fully detects sleep apnea for that person. The system also healthy and sound life without a sleeping disorder is a bless-
detects obstructive sleep apnea in a person. After analyzing ing, and technology can undoubtedly assure that.
the results, it is clear that the person whose age is 50+ is a
patient with OSA. That kind of monitoring will help people
to detect sleep apnea at the early stage. Thus, this research Data Availability
can help people to learn about sleep apnea, the way to detect
it, and it will also help people to eliminate all their sleeping No data was utilized to support this research findings.
problems.
Nowadays, sleeping disorder rates are higher than
before. More people are suffering from hypertension, Conflicts of Interest
stress-related problems, and heart-related problems which The authors declare that they have no conflicts of interest to
cause sleep apnea. Our system will be very much more sus- report regarding the present study.
tainable to monitor sleep apnea and detect specifically the
duration of sleep apnea. That system can sustain itself at a
higher level. It not only detects sleep apnea but can also Acknowledgments
inform the patient to discuss with the doctor the data taken
from the system, which can make the lives of patients easier. The authors are thankful for the support from Taif Univer-
And the project can be sustained in the future if we add sity Researchers Supporting Project (TURSP-2020/26), Taif
higher technology like high-level X-rays and other monitor- University, Taif, Saudi Arabia.
16 Computational and Mathematical Methods in Medicine

References [16] C. Gouveia, H. Qureshi, R. Kern, S. Liu, and R. Capasso, “An

assessment of online information related to surgical obstruc-
[1] T. Kasai, “Sleep apnea and heart failure,” Journal of Cardiology, tive sleep apnea treatment,” Sleep Science and Practice, vol. 1,
vol. 60, no. 2, pp. 78–85, 2012. no. 1, pp. 1–6, 2017.
[2] J. Jin and E. Sánchez-Sinencio, “A home sleep apnea screening [17] J. P. Kirkness, B. M. McGinley, F. P. Sgambati et al., “Develop-
device with time-domain signal processing and autonomous ing quantitative physiological phenotypes of sleep apnea for
scoring capability,” IEEE Transactions on Biomedical Circuits epidemiological studies,” in Annual International Conference
and Systems, vol. 9, no. 1, pp. 96–104, 2015. of the IEEE Engineering in Medicine and Biology Society,
[3] A. Malhotra and D. P. White, “Obstructive sleep apnoea,” Lan- pp. 8319–8322, Boston, MA, USA, 2011.
cet, vol. 360, no. 9328, pp. 237–245, 2002. [18] “Arduino Uno, components101,” https://components101
[4] J. Crow, “Sleep Apnea Statistics. Rest Right Mattress,” https:// .com/microcontrollers/arduino-uno.
restrightmattress.com/sleep-apnea-statistics/. [19] A. Goldberger, “Giles F. Filley lecture. Complex systems,” Pro-
[5] M. Rakicevic, 31 Important Sleep Apnea Statistics You Should ceedings of the American Thoracic Society, vol. 3, no. 6,
Know in 2021, Disturbmenothttps://disturbmenot.co/sleep- pp. 467–471, 2016.
apnea-statistics. [20] M. Al-Mardini, F. Aloul, A. Sagahyroon, and L. Al-Husseini,
[6] R. Boneberg, A. Pardun, L. Hannemann et al., “High plasma “Classifying obstructive sleep apnea using smartphones,” Jour-
cystine levels are associated with blood pressure and reversed nal of Biomedical Informatics, vol. 52, pp. 251–259, 2014.
by CPAP in patients with obstructive sleep apnea,” Journal of [21] “Max30100 pulse oximeter Arduino code, circuit, and pro-
Clinical Medicine, vol. 10, no. 7, p. 1387, 2021. gramming, electronic clinic,” https://www.electroniclinic
[7] G. Cay and K. Mankodiya, “SleepSmart: smart mattress inte- .com/max30100-pulse-oximeter-arduino-code-circuit-and-
grated with e-textiles and IoT functions for sleep apnea man- programming/.
agement,” in 2020 IEEE International Conference on [22] P. A. Indursky, “The phenomenon of weak, unawakening
Pervasive Computing and Communications Workshops (Per- afferent signals involvement in synchronization mechanisms
Com Workshops), pp. 1-2, Austin, USA, 2020. of neuron activity during delta sleep-unawakening afferent sig-
[8] I. Sadek, E. Seet, J. Biswas, B. Abdulrazak, and M. Mokhtari, nals involvement in synchronization mechanisms of neuron
“Nonintrusive vital signs monitoring for sleep apnea patients: activity during delta sleep,” International Journal of Sleep Dis-
a preliminary study,” IEEE Access, vol. 6, pp. 2506–2514, 2018. orders, vol. 1, no. 1, pp. 8–11, 2017.
[9] Su Hwan Hwang, Hong Ji Lee, Hee Nam Yoon et al., “Uncon- [23] “Mit App Inventor 2,” http://appinventor.mit.edu.
strained sleep apnea monitoring using polyvinylidene fluoride [24] “Circuit design app for makers- circuito.io, circuit design app
film-based sensor,” IEEE Transactions on Biomedical Engi- for makers- circuito.io,” https://www.circuito.io/.
neering, vol. 61, no. 7, pp. 2125–2134, 2014. [25] G. Lagana, N. Venza, A. Malara, C. Liguori, P. Cozza, and
[10] T. Penzel, M. Glos, I. Fietze, S. Herberger, and G. Pillar, “Dis- C. Pisano, “Obstructive sleep apnea, palatal morphology, and
tinguish obstructive and central sleep apnea by portable aortic dilatation in Marfan syndrome growing subjects: a ret-
peripheral arterial tonometry,” in 42nd Annual International rospective study,” International Journal of Environmental
Conference of the IEEE Engineering in Medicine & Biology Soci- Research and Public Health, vol. 18, no. 6, p. 3045, 2021.
ety (EMBC), pp. 2780–2783, Montreal, QC, Canada, 2020. [26] B. L. Koley and D. Dey, “Real-time adaptive apnea and hypop-
[11] D. Falie and M. Ichim, “Sleep monitoring and sleep apnea nea event detection methodology for portable sleep apnea
event detection using a 3D camera,” in 2010 8th International monitoring devices,” IEEE Transactions on Biomedical Engi-
Conference on Communications, pp. 177–180, Bucharest, neering, vol. 60, no. 12, pp. 3354–3363, 2013.
Romania, 2010. [27] L. A. Lee, Y. L. Lo, J. F. Yu et al., “Snoring sounds predict
[12] E. Dafna, A. Tarasiuk, and Y. Zigel, “Sleep-quality assessment obstruction sites and surgical response in patients with
from full night audio recordings of sleep apnea patients,” in obstructive sleep apnea hypopnea syndrome,” Scientific
2012 Annual International Conference of the IEEE Engineering Reports, vol. 6, no. 1, 2016.
in Medicine and Biology Society, pp. 3660–3663, San Diego,
CA, USA, 2012.
[13] T. Reza, S. B. A. Shoilee, S. M. Akhand, and M. M. Khan,
“Development of android based pulse monitoring system,” in
2017 Second International Conference on Electrical, Computer
and Communication Technologies (ICECCT), pp. 1–7, Coim-
batore, India, 2017.
[14] L. Almazaydeh, K. Elleithy, and M. Faezipour, “Obstructive
sleep apnea detection using SVM-based classification of ECG
signal features,” in 2012 Annual International Conference of
the IEEE Engineering in Medicine and Biology Society,
pp. 4938–4941, San Diego, CA, USA, 2012.
[15] L. Cen, Z. L. Yu, T. Kluge, and W. Ser, “Automatic system for
obstructive sleep apnea events detection using convolutional
neural network,” in 2018 40th Annual International Confer-
ence of the IEEE Engineering in Medicine and Biology Society
(EMBC), pp. 3975–3978, Honolulu, HI, USA, 2018.