SLEEPING ABNORMALITIES DETECTION USING DEEP

LEARNING TECHNIQUE

A PROJECT REPORT

Submitted by

ANANTHANAYAKI.S 822219104002

SANTHOSH.S 822219104033

SWATHIPRIYA.K 822219104037

in partial fulfilment for the award of the degree

of

BACHELOR OF ENGINEERING

IN

COMPUTER SCIENCE & ENGINEERING

UNIVERSITY COLLEGE OF ENGINEERING THIRUKKUVALAI

(A Constituent College of Anna University, Chennai)

ANNA UNIVERSITY: CHENNAI 600 025

MAY 2023

I
 ANNA UNIVERSITY: CHENNAI 600025

BONAFIDE CERTIFICATE

Certified that this project report titled “SLEEPING ABNORMALITIES

DETECTION USING DEEP LEARNING TECHNIQUE” is the bonafide work of
“ANANTHANAYAKI S(822219104002), SANTHOSH S (822219104033)”,
SWATHIPRIYA K (822219104037)” of Computer Science & Engineering whose
carried out this project work under my supervision.

SIGNATURE SIGNATURE
Dr.K.L.NEELA,M.E., Ph.D., Mr.K.SATHEESHKUMAR,MCA., M.Phil.,
Assistant professor, Assistant professor,
HEAD OF THE DEPARTMENT, SUPERVISOR,
Department of CSE, Department of CSE,
University College of Engineering, University College of Engineering,
Thirukkuvalai-610204. Thirukkuvalai-610204.

Submitted for the Project Viva Voice held on: ……………….

INTERNAL EXAMINER EXTERNAL EXAMINER

II
 ACKNOWLEDGEMENT

We are always thankful to Dr. G. ELANGOVAN, M.E, Ph.D., DEAN,

University College of Engineering, Thirukkuvalai, who endorsed us throughout
this project.

We express our sincere thanks to Head of the Department

Dr.K.L.NEELA,M.E.,Ph.D, for providing the necessary support in all means in
carry out this project.

We express our sincere thanks to the project committee co-ordinator

Mr.S.MADHAN ,B.E,M.TECH., Assistant Psrofessor, Department of Computer
Science and Engineering, University College ofEngineering, Thirukkuvalai, for
their invaluable guidance and technical Support. Wethank all the faculty members
of the Department of Computer Science and Engineering for their valuable help
and guidance.

We express our deep sense of reverence and offer our thanks with profound
gratitude to our project guide Mr.K.SATHEESHKUMAR,MCA.,M.Phil, for
motivation and encourage infused in us. Hes readiness and eagerness for
consolation at all times, him educative inputs, concern and assistance have been
invaluable.

Finally we thank god almighty for his blessing without which we would have not
the project and made this project success.

ANANTHANAYAKI.S SANTHOSH.S SWATHIPRIYA.K

822219104002 822219104033 822219104037

III
 ABSTRACT

Maintaining proper health and mental stability is critical for overall health and
well-being. Despite a good deal of research investment, sleep quality continues to
be a crucial public challenge. Nowadays, people of all age groups are affected by
improper sleep quality. Poor sleep can lead to a variety of neurological disorders.
Sleep disorders are common in all subsets of the population, independently of
gender. This public health challenge greatly affects quality of life in terms of both
physical and mental health. Insomnia, parasomnias, sleep-related breathing
difficulties, hypersomnia, bruxism, narcolepsy, and circadian rhythm disorders are
some common examples of sleep-related disorders. Some of these disorders can be
treated with proper analysis of early symptoms; in such cases, adequate sleep quality
is essential for the patient’s recovery. Artificial intelligence has several applications
in sleep medicine including sleep and respiratory event scoring in the sleep
laboratory, diagnosing and managing sleep disorders, and population health. While
still in its nascent stage, there are several challenges which preclude AI’s
generalizability and wide-reaching clinical applications. Artificial intelligence is a
powerful tool in healthcare that may improve patient care, enhance diagnostic
abilities, and augment the management of sleep disorders. However, there is a need
to regulate and standardize existing machine learning algorithms and deep learning
algorithm prior to its inclusion in the sleep clinic. In this project, we can develop the
framework for sleeping stage classification for both subjective data and ECG data
and classify the data using Convolutional neural network algorithm to analyse the
multiple sleeping stages.

IV
 TABLE OF CONTENTS

CHAPTER TITLE PAGE

NO NO

ABSTRACT IV
LIST OF FIGURES VIII
LIST OF ABBREVIATION IX
1 INTRODUCTION 1
1.1 DEEP LEARNING 1
1.2 APPLICATIONS OF DEEP LEARNING 2
1.3 DEEP LEARNING WORKS 2
1.4 PREDICTIVE ANALYSIS 5
1.5 MACHINE LEARNING 7
1.6 ARTIFICIAL INTELLIGENCE 9
1.7 REINFORCEMENT LEARNING 11

2 LITERATURE REVIEW 14
2.1 DETECTION OF INSOMNIA USING ECG AND
EMG 14
2.2 AUTOMATIC IDENTIFICATION OF
INSOMNIA BASED ON SINGLE EEG 15
2.3 A SHORT TIME INSOMNIA 16
2.4 INSOMNIA TOWARD CONCEPT DRIFT
ROBUSTNESS 17
2.5 AUTOMATIC IDENTIFICATION OF
INSOMNIA USING OPTIMAL ANTISYMMETRIC 18
2.6 MULTITASK FMRI AND MACHINE
19
LEARNING APPROACH
2.7 DISCRIMINATING PARADOXICAL AND
20
PSYCHOPHSIOLOGIAL

V
3 SYSTEM ANALYSIS 21
3.1 EXISTING SYSTEM 21
3.1.1 DISADVANTAGES 22
3.2 PROPOSED SYSTEM 22
3.3 CONVOLUTIONAL NEURAL NETWORK 23
3.3.1 APPLICATION 24
25
3.3.2 ADVANTAGES

4 SYSTEM IMPLEMENTATION 26
4.1 DATASETS 26
4.2 HARDWARE REQUIRMENTS 28
4.3 SOFTWARE REQUIRMENTS 28
4.4 MODULES 28
4.4.1 DATASETS ACQUSITION 28
4.4.2 PREPROCESSING 29
4.4.3 FEATURS EXTRACTION 30
4.4.4 CLASSIFICATION 31
4.4.5 PREDICTION 32

5 SYSTEM DESIGN 33
5.1 SYSTEM ARCHITECTURE 33
5.2 UML DIAGRAMS 34
5.2.1 USE CASE DIAGRAM 34
5.2.2 CLASS DIAGRAM 35
5.2.3 ACTIVITY DIAGRAM 35
5.2.4 SEQUENCE DIAGRAM 36
5.2.5 COLLABORATION DIAGRAM 37

6 SYSTEM TESTING 39
6.1 SOFTWARE TESTING STRATEGIES 39
6.1.1 UNIT TESTING 39
6.1.2 FUNCTIONAL TESTING 40
6.1.3 ACCEPTANCE TESTING 40

7 RESULT AND ANALYSIS 41

VI
8 SNAPSHOTS 42

9 APPENDIX 53

10 CONCLUSION AND FUTURE WORK 66

11 REFERENCES 67

12 RESEARCH WORK PUBLICATION 70

VII
 LIST OF FIGURES

FIGURE NO NAME OF THE FIGURE PAGE NO

5.1 SYSTEM ARCHITECTURE 33
5.2 UML DIAGRAMS 34
5.2.1 USE CASE DIAGRAM 34
5.2.2 CLASS DIAGRAM 35
5.2.3 ACTIVITY DIAGRAM 35
5.2.4 SEQUENCE DIAGRAM 36
5.2.5 COLLABORATION DIAGRAM 37

VIII
 LIST OF ABBREVIATIONS

S.NO ABBREVIATIONS EXPANSION

1 EEG ELECTROENCEPHALOGRAM

2 ECG ELECTROCARDIOGAPHY

3 EMG ELECTROMYOGRAPHY

IX
 CHAPTER 1
INTRODUCTION

1.1 DEEP LEARNING

Deep learning is a machine learning technique that teaches computers to do

what comes naturally to humans: learn by example. Deep learning is a key
technology behind driverless cars, enabling them to recognize a stop sign, or to
distinguish a pedestrian from a lamppost. It is the key to voice control in consumer
devices like phones, tablets, TVs, and hands-free speakers. Deep learning is getting
lots of attention lately and for good reason. It’s achieving results that were not
possible before. In deep learning, a computer model learns to perform classification
tasks directly from images, text, or sound. Deep learning models can achieve state-
of-the-art accuracy, sometimes exceeding human-level performance. Models are
trained by using a large set of labeled data and neural network architectures that
contain many layers. Deep learning achieves recognition accuracy at higher levels
than ever before. This helps consumer electronics meet user expectations, and it is
crucial for safety-critical applications like driverless cars. Recent advances in deep
learning have improved to the point where deep learning outperforms humans in
some tasks like classifying objects in images. While deep learning was first
theorized in the 1980s, there are two main reasons it has only recently become
useful: Deep learning requires large amounts of labeled data. For example,
driverless car development requires millions of images and thousands of hours of
video. Deep learning requires substantial computing power. High-performance
GPUs have a parallel architecture that is efficient for deep learning. When
combined with clusters or cloud computing, this enables development teams to
reduce training time for a deep learning network from weeks to hours or less.

1
1.2 APPLICATIONS OF DEEP LEARNING
Deep learning applications are used in industries from automated driving to
medical devices.Automated Driving: Automotive researchers are using deep
learning to automatically detect objects such as stop signs and traffic lights. In
addition, deep learning is used to detect pedestrians, which helps decrease
accidents.

Aerospace and Defense: Deep learning is used to identify objects from satellites
that locate areas of interest, and identify safe or unsafe zones for troops.

Medical Research: Cancer researchers are using deep learning to automatically

detect cancer cells. Teams at UCLA built an advanced microscope that yields a
high-dimensional data set used to train a deep learning application to accurately
identify cancer cells.

Industrial Automation: Deep learning is helping to improve worker safety around

heavy machinery by automatically detecting when people or objects are within an
unsafe distance of machines.

Electronics: Deep learning is being used in automated hearing and speech

translation. For example, home assistance devices that respond to your voice and
know your preferences are powered by deep learning applications.

1.3 DEEP LEARNING WORKS

Most deep learning methods use neural network architectures, which is why
deep learning models are often referred to as deep neural networks.

The term “deep” usually refers to the number of hidden layers in the neural
network. Traditional neural networks only contain 2-3 hidden layers, while deep
networks can have as many as 150.

2
Deep learning models are trained by using large sets of labeled data and neural
network architectures that learn features directly from the data without the need for
manual feature extraction.

One of the most popular types of deep neural networks is known as convolutional
neural networks (CNN or ConvNet). A CNN convolves learned features with
input data, and uses 2D convolutional layers, making this architecture well suited
to processing 2D data, such as images.

CNNs eliminate the need for manual feature extraction, so you do not need to
identify features used to classify images. The CNN works by extracting features
directly from images. The relevant features are not pretrained; they are learned
while the network trains on a collection of images. This automated feature
extraction makes deep learning models highly accurate for computer vision tasks
such as object classification.

CNNs learn to detect different features of an image using tens or hundreds of

hidden layers. Every hidden layer increases the complexity of the learned image
features. For example, the first hidden layer could learn how to detect edges, and
the last learns how to detect more complex shapes specifically catered to the shape
of the object we are trying to recognize.

Deep learning is a specialized form of machine learning. A machine learning

workflow starts with relevant features being manually extracted from images. The
features are then used to create a model that categorizes the objects in the image.
With a deep learning workflow, relevant features are automatically extracted from
images. In addition, deep learning performs “end-to-end learning” – where a
network is given raw data and a task to perform, such as classification, and it learns
how to do this automatically.

3
Another key difference is deep learning algorithms scale with data, whereas
shallow learning converges. Shallow learning refers to machine learning methods
that plateau at a certain level of performance when you add more examples and
training data to the network.

A key advantage of deep learning networks is that they often continue to improve
as the size of your data increases.

The three most common ways people use deep learning to perform object
classification are:

Training from Scratch

To train a deep network from scratch, you gather a very large labeled data set
and design a network architecture that will learn the features and model. This is
good for new applications, or applications that will have a large number of output
categories. This is a less common approach because with the large amount of data
and rate of learning, these networks typically take days or weeks to train.

Transfer Learning

Most deep learning applications use the transfer learning approach, a process
that involves fine-tuning a pretrained model. You start with an existing network,
such as AlexNet or GoogLeNet, and feed in new data containing previously
unknown classes. After making some tweaks to the network, you can now perform
a new task, such as categorizing only dogs or cats instead of 1000 different objects.
This also has the advantage of needing much less data (processing thousands of
images, rather than millions), so computation time drops to minutes or hours.

4
Feature Extraction

A slightly less common, more specialized approach to deep learning is to use

the network as a feature extractor. Since all the layers are tasked with learning
certain features from images, we can pull these features out of the network at any
time during the training process. These features can then be used as input to
a machine learning model such as support vector machines (SVM).

1.4 PREDICTIVE ANALYSIS

Predictive analytics uses historical data to predict future events. Typically,

historical data is used to build a mathematical model that captures important trends.
That predictive model is then used on current data to predict what will happen next,
or to suggest actions to take for optimal outcomes.

Predictive analytics has received a lot of attention in recent years due to advances
in supporting technology, particularly in the areas of big data and machine learning.

Predictive analytics is often discussed in the context of big data, Engineering data,
for example, comes from sensors, instruments, and connected systems out in the
world. Business system data at a company might include transaction data, sales
results, customer complaints, and marketing information. Increasingly, businesses
make data-driven decisions based on this valuable trove of information. To extract
value from big data, businesses apply algorithms to large data sets using tools such
as Hadoop and Spark. The data sources might consist of transactional databases,
equipment log files, images, video, audio, sensor, or other types of data. Innovation
often comes from combining data from several sources.

With all this data, tools are necessary to extract insights and trends. Machine
learning techniques are used to find patterns in data and to build models that predict
future outcomes. A variety of machine learning algorithms are available, including

5
linear and nonlinear regression, neural networks, support vector machines, decision
trees, and other algorithms.

Predictive analytics helps teams in industries as diverse as finance, healthcare,

pharmaceuticals, automotive, aerospace, and manufacturing.

• Automotive – Breaking new ground with autonomous vehicles

o Companies developing driver assistance technology and new
autonomous vehicles use predictive analytics to analyze sensor data from
connected vehicles and to build driver assistance algorithms.
• Aerospace – Monitoring aircraft engine health
o To improve aircraft up-time and reduce maintenance costs, an engine
manufacturer created a real-time analytics application to predict
subsystem performance for oil, fuel, liftoff, mechanical health, and
controls.
• Energy Production – Forecasting electricity price and demand
o Sophisticated forecasting apps use models that monitor plant availability,
historical trends, seasonality, and weather.
• Financial Services – Developing credit risk models
o Financial institutions use machine learning techniques and quantitative
tools to predict credit risk.
• Industrial Automation and Machinery – Predicting machine failures
o A plastic and thin film producer saves 50,000 Euros monthly using a
health monitoring and predictive maintenance application that reduces
downtime and minimizes waste.
• Medical Devices – Using pattern-detection algorithms to spot asthma and
COPD. An asthma management device records and analyzes patients' breathing

6
 sounds and provides instant feedback via a smart phone app to help patients
manage asthma and COPD.

• Predictive analytics– It is the process of using data analytics to make

predictions based on data. This process uses data along with analysis, statistics,
and machine learning techniques to create a predictive model for forecasting
future events.

The term “predictive analytics” describes the application of a statistical or machine

learning technique to create a quantitative prediction about the future.
Frequently, supervised machine learning techniques are used to predict a future
value (How long can this machine run before requiring maintenance?) or to
estimate a probability (How likely is this customer to default on a loan?). Predictive
analytics starts with a business goal: to use data to reduce waste, save time, or cut
costs. The process harnesses heterogeneous, often massive, data sets into models
that can generate clear, actionable outcomes to support achieving that goal, such as
less material waste, less stocked inventory, and manufactured product that meets
specifications.

1.5 MACHINE LEARNING

Machine learning is a data analytics technique that teaches computers to do

what comes naturally to humans and animals: learn from experience. Machine
learning algorithms use computational methods to “learn” information directly
from data without relying on a predetermined equation as a model. The algorithms
adaptively improve their performance as the number of samples available for
learning increases. Deep learning is a specialized form of machine learning. With
the rise in big data, machine learning has become a key technique for solving
problems in areas, such as:

7
 • Computational finance, for credit scoring and algorithmic trading
• Image processing and computer vision, for face recognition, motion
detection, and object detection
• Computational biology, for tumor detection, drug discovery, and DNA
sequencing
• Energy production, for price and load forecasting
• Automotive, aerospace, and manufacturing, for predictive maintenance
• Natural language processing, for voice recognition applications

Machine learning uses two types of techniques: supervised learning, which trains
a model on known input and output data so that it can predict future outputs,
and unsupervised learning, which finds hidden patterns or intrinsic structures in
input data.

Supervised Learning

Supervised machine learning builds a model that makes predictions based on

evidence in the presence of uncertainty. A supervised learning algorithm takes a
known set of input data and known responses to the data (output) and trains a model
to generate reasonable predictions for the response to new data. Use supervised
learning if you have known data for the output you are trying to predict. Supervised
learning uses classification and regression techniques to develop predictive
models.

Classification Techniques

Predict discrete responses for example, whether an email is genuine or spam,

or whether a tumor is cancerous or benign. Classification models classify input
data into categories. Typical applications include medical imaging, speech
recognition, and credit scoring. Use classification if your data can be tagged,
categorized, or separated into specific groups or classes. For example, applications

8
for hand-writing recognition use classification to recognize letters and numbers. In
image processing and computer vision, unsupervised pattern
recognition techniques are used for object detection and image segmentation.

Regression Techniques

Predict continuous responses for example, changes in temperature or

fluctuations in power demand. Typical applications include electricity load
forecasting and algorithmic trading.

Use regression techniques if you are working with a data range or if the nature of
your response is a real number, such as temperature or the time until failure for a
piece of equipment.

Unsupervised Learning

Unsupervised learning finds hidden patterns or intrinsic structures in data. It

is used to draw inferences from datasets consisting of input data without labeled
responses.

Clustering

It is the most common unsupervised learning technique. It is used for

exploratory data analysis to find hidden patterns or groupings in data. Applications
for cluster analysis include gene sequence analysis, market research, and object
recognition.

1.6 ARTIFICIAL INTELLIGENCE

Artificial intelligence, or AI, is a simulation of intelligent human behavior.

It’s a computer or system designed to perceive its environment, understand its
behaviors, and take action. Consider self-driving cars: AI-driven systems like these
9
integrate AI algorithms, such as machine learning and deep learning, into complex
environments that enable automation. Data Preparation

Taking raw data and making it useful for an accurate, efficient, and meaningful
model is a critical step. In fact, it represents most of your AI effort.

Data preparation requires domain expertise, such as experience in speech and audio
signals, navigation and sensor fusion, image and video processing, and radar and
lidar. Engineers in these fields are best suited to determine what the critical features
of the data are, which are unimportant, and what rare events to consider.

AI also involves prodigious amounts of data. Yet labeling data and images
is tedious and time-consuming. Sometimes, you don’t have enough data, especially
for safety-critical systems. Generating accurate synthetic data can improve your
data sets. In both cases, automation is critical to meeting deadlines.

Deployment

AI models need to be deployed to CPUs, GPUs, and/or FPGAs in your final

product, whether part of an embedded or edge device, enterprise system, or cloud.
AI models running on the embedded or edge device provide the quick results
needed in the field, while AI models running in enterprise systems and the cloud
provide results from data collected across many devices. Frequently, AI models are
deployed to a combination of these systems.

The deployment process is accelerated when you generate code from your models
and target your devices. Using code generation optimization techniques and
hardware-optimized libraries, you can tune the code to fit the low power profile
required by embedded and edge devices or the high-performance needs of
enterprise systems and the cloud.

10
1.7 REINFORCEMENT LEARNING

In control systems that benefit from learning based on cumulative reward,

reinforcement learning is an ideal technique. Reinforcement Learning
Toolbox™ lets you train policies using DQN, A2C, DDPG, and other
reinforcement learning algorithms. You can use these policies to implement
controllers and decision-making algorithms for complex systems such as robots
and autonomous systems. You can implement the policies using deep neural
networks, polynomials, or lookup tables. Reinforcement learning is a type of
machine learning technique where a computer agent learns to perform a task
through repeated trial and error interactions with a dynamic environment. This
learning approach enables the agent to make a series of decisions that maximize a
reward metric for the task without human intervention and without being explicitly
programmed to achieve the task. AI programs trained with reinforcement learning
beat human players in board games like Go and chess, as well as video games.
While reinforcement learning is by no means a new concept, recent progress in
deep lerning and computing power made it possible to achieve some remarkable
results in the area of artificial intelligence. Reinforcement learning is a branch of
machine learning (Figure 1). Unlike unsupervised and supervised machine
learning, reinforcement learning does not rely on a static dataset, but operates in a
dynamic environment and learns from collected experiences. Data points, or
experiences, are collected during training through trial-and-error interactions
between the environment and a software agent. This aspect of reinforcement
learning is important, because it alleviates the need for data collection,
preprocessing, and labeling before training, otherwise necessary in supervised and
unsupervised learning. Practically, this means that, given the right incentive, a
reinforcement learning model can start learning a behavior on its own, without
(human) supervision.

11
Deep learning spans all three types of machine learning; reinforcement learning
and deep learning are not mutually exclusive. Complex reinforcement learning
problems often rely on deep neural networks, a field known as deep reinforcement
learning

Deep neural networks trained with reinforcement learning can encode complex
behaviors. This allows an alternative approach to applications that are otherwise
intractable or more challenging to tackle with more traditional methods. For
example, in autonomous driving, a neural network can replace the driver and
decide how to turn the steering wheel by simultaneously looking at multiple
sensors such as camera frames and lidar measurements. Without neural networks,
the problem would normally be broken down in smaller pieces like extracting
features from camera frames, filtering the lidar measurements, fusing the sensor
outputs, and making “driving” decisions based on sensor inputs.

While reinforcement learning as an approach is still under evaluation for

production systems, some industrial applications are good candidates for this
technology.

Advanced controls: Controlling nonlinear systems is a challenging problem that

is often addressed by linearizing the system at different operating points.
Reinforcement learning can be applied directly to the nonlinear system.

Automated driving: Making driving decisions based on camera input is an area

where reinforcement learning is suitable considering the success of deep neural
networks in image applications.

Robotics: Reinforcement learning can help with applications like robotic grasping,
such as teaching a robotic arm how to manipulate a variety of objects for pick-and-

12
place applications. Other robotics applications include human-robot and robot-
robot collaboration.

Scheduling: Scheduling problems appear in many scenarios including traffic light

control and coordinating resources on the factory floor towards some objective.
Reinforcement learning is a good alternative to evolutionary methods to solve these
combinatorial optimization problems.

Calibration: Applications that involve manual calibration of parameters, such as

electronic control unit (ECU) calibration, may be good candidates for
reinforcement learning.

13
 CHAPTER 2
LITERATURE SURVEY

2.1 TITLE: DETECTION OF INSOMNIA USING ECG AND EMG

AUTHOR: ASRA MALIK YEAR 2021

Insomnia is a disorder of sleep in which a person constantly complains about

insufficient sleep despite having an adequate amount of time to sleep. It is
considered as one of the most common psychological illnesses with an
approximate severity of about 10%. Frustration is the key feature of insomnia. The
quality and quantity of sleep remain in trouble while falling asleep, sustaining
sleep, or rising in the early morning. Insomnia is also associated with several other
sleep-wake, emotional, or medical conditions which rises attention for its separate
treatment. Increased physical, psychological, and physiological stimulation is
considered to be a key underlying in most reports of chronic insomnia along with
perpetuating behavioural issues like prolonged time in bed. It is also reported in a
research work that insomnia can also combine with cardiovascular diseases. It has
a very adverse effect on the job, mental, social, and overall quality of life of an
individual. Insomnia is one of the serious sleep disorders. In this disease, a patient
feels trouble falling asleep and stay awake for the whole night. Insomnia also co-
morbid with other diseases like cancer, cardiovascular disease, asthma,
Alzheimer's, Parkinson's disease. This research aims at a method for identifying
insomnia using both electrocardiogram (ECG) and electromyogram (EMG)
signals. The Cyclic Alternating Pattern (CAP) database from Physionet is used.

14
2.2 TITLE: AUTOMATIC IDENTIFICATION OF INSOMNIA
BASED ON SINGLE-CHANNELEEG LABELLED WITH SLEEP
STAGE ANNOTATIONS
AUTHOR: BUFANG YANG YEAR 2020

In this paper, we proposed a 1D-CNN model for automatic insomnia

identification based on single-channel EEG labelled with sleep stage annotations,
and further investigated the identification performance based on different sleep
stages. Our experiments demonstrated that our 1D-CNN leveraging the 3 sub
datasets composed of REM, LSS and SWS epochs, respectively, achieved higher
average accuracy in comparison with baseline methods under both intra-patient and
inter-patient paradigms. The experimental results also indicated that amongst all
the sleep stages, 1D-CNN leveraging REM and SWS epochs exhibited the best
insomnia identification performance in intra-patient paradigm, whereas no
statistically significant difference was found in inter-patient paradigm. Overall, for
automatic insomnia identification based on single-channel EEG labelled with sleep
stages, 1D-CNN model introduced in this paper could achieve superior
performance than traditional methods. Further experiment based on larger sleep
databases under inter-patient paradigm is still required in future work. Moreover,
the comparison of the two experiments demonstrated that the average identification
accuracy of our 1D-CNN could achieve 99.16% in intra-patient experiment,
whereas it could only reach 87.49% in inter-patient experiment with larger standard
deviation. We consider the high accuracy in intra-patient experiment is caused by
the similarity of epochs, i.e. epochs from the same patient are utilized both for
training and testing. However, inter-patient experiment is more realistic evaluation
paradigm which could guarantee the generalizability of the method.

15
2.3 TITLE: A SHORT-TIME INSOMNIA DETECTION SYSTEM
BASED ON SLEEP EOG WITH RCMSE ANALYSIS
AUTHOR: CHIH-EN KUO YEAR 2020

Sleep takes approximately one-third of human live. A good sleep can help us
getting the body to work right again, improved learning ability, physical
development, emotional regulation, and good quality of life in human physiology.
However, the prevalence of insomnia symptoms without restrictive criteria is
approximately 33% in the general population. In the United States of America, 50-
70 million people suffer from sleep disorders: among them, 30% of patients suffer
from insomnia and 10% from chronic insomnia. Insomnia is defined as chronic
when it has persisted for at least three months at a frequency of at least three times
per week. When the disorder meets the symptom criteria but has persisted for less
than three months, it is considered short-term insomnia. To diagnose insomnia, the
physician may first execute a physical exam to look for signs of medical problems
that may be related to insomnia and ask some sleep-related questions, such as
sleep-wake pattern and daytime sleepiness. In addition, the physician may ask a
subject to keep a sleep diary with the actigraphy for a couple of weeks. If the cause
of insomnia is not clear, the subject should spend one or two nights at a sleep centre
diagnosing another sleep disorder using polysomnography (PSG), such as sleep
apnea. PSG recordings, which including electroencephalogram (EEG),
electrooculogram (EOG), electromyogram (EMG), and other physiological signals
are usually obtained from patients and scored by a well-trained clinical staff.
Moreover, manual sleep scoring and diagnosis is a time consuming and subjective
process, and the conventional polysomnography which uses many wires to connect
instrument to patient is often a problem that leads to sleep disturbance.

16
2.4 TITLE: INSOMNIA: TOWARDS CONCEPT-DRIFT
ROBUSTNESS IN NETWORK INTRUSION DETECTION
AUTHOR: GIUSEPPINA ANDRESINI YEAR 2021

This paper outlines a set of open challenges facing modern ML-based

intrusion detectors relating to a lack of uniformity in the distribution of traffic data
over time. To tackle them, we propose INSOMNIA, a semi-supervised approach
that uses active learning to reduce latency in the model updates, label estimation to
reduce labelling overhead, and applies explainable AI to describe how the model
evolves to fit the shifting distribution. We extend the TESSERACT framework to
perform a time-aware evaluation of INSOMNIA on a recently published, revised
version of CICIDS2017 and demonstrate that modern intrusion detection systems
must address concept drift in order to be effective. We envision that future work
may build on INSOMNIA in order to design robust intrusion detection models that
can be sustained over time. They rely on a subset of ground truth labels to retrain
at each time step. In contrast to these previously mentioned approaches,
INSOMNIA only requires labels at the initial training time, and then sustains itself
without requiring manual labelling; instead, it generates pseudo labels based on the
nearest centroid neighbour. Updating with the model’s predicted labels has been
proposed as a solution in the malware domain, but has been shown to lead to self-
poisoning; this risk is mitigated in INSOMNIA’s use of curtaining by using two
distinct algorithms for label estimation and prediction. The results of our evaluation
highlight yet another important open problem for NIDS: detection of stealthy, low-
prevalence attacks. INSOMNIA struggles to detect the few instances of the
Infiltration attack at windows and we observe that maintaining sensitivity to attacks
with a very low base rate in the presence of high-volume attacks such as DoS is
very challenging—as is generalizing to attack categories of greatly different
character.

17
2.5 TITLE: AUTOMATIC IDENTIFICATION OF INSOMNIA
USING OPTIMAL ANTISYMMETRIC BIORTHOGONAL
WAVELET FILTER BANK WITH ECG SIGNALS
AUTHOR: MANISH SHARMA YEAR 2021

Sleep is a fundamental human physiological activity required for adequate

working of the human body. Sleep disorders such as sleep movement disorders,
nocturnal front lobe epilepsy, insomnia, and narcolepsy are caused due to low sleep
quality. Insomnia is one such sleep disorder where a person has difficulty in getting
quality sleep. There is no definitive test to identify insomnia; hence it is essential
to develop an automated system to identify it accurately. A few automated methods
have been proposed to identify insomnia using either polysomnogram (PSG) or
electroencephalogram (EEG) signals. To the best of our knowledge, we are the first
to automatically detect insomnia using only electrocardiogram (ECG) signals
without combining them with any other physiological signals. In the proposed
study, an optimal antisymmetric biorthogonal wavelet filter bank (ABWFB) has
been used, which is designed to minimize the joint duration-bandwidth localization
(JDBL) of the underlying filters. We created ten different subsets of ECG signals
based on annotations of sleepstages, namely wake (W), S1, S2, S3, S4, rapid eye
moment (REM), light sleep stage (LSS), slow-wave sleep (SWS), non-rapid eye
movement (NREM) and W+S1+S2+S3+S4+REM for the automated identification
of insomnia. Our proposed ECG-based system obtained the highest classification
accuracy of 97.87%, F1-score of 97.39%, and Cohen’s kappa value of 0.9559 for
K-nearest neighbour (KNN) with the ten-fold cross-validation strategy using ECG
signals corresponding to the REM sleep stage. The support vector machine (SVM)
yielded the highest value of 0.99 for area under the curve with the tenfold cross-
validation corresponding to REM sleep stage

18
2.6 TITLE: MULTITASK FMRI AND MACHINE LEARNING
APPROACH IMPROVE PREDICTION OF DIFERENTIAL
BRAIN ACTIVITY PATTERN IN PATIENTS WITH INSOMNIA
DISORDER
AUTHOR: MI HYUN LEE YEAR 2021

Insomnia is a common, distressing, and clinically important symptom, which

causes difficulty initiating sleep; frequent awakening; or early-morning awakening
with daytime dysfunction. Approximately one-third of the general population
experiences insomnia symptoms throughout their lifetime; 1-year follow-up of
patients with insomnia revealed that 69% exhibited chronic symptoms. Because of
its high prevalence and frequent associations with medical and psychiatric
disorders, there is a need to study the neurobiological mechanisms of insomnia to
improve quality of life and reduce socioeconomic burden. According to the
International Classification of Sleep Disorders Second Edition (ICSD-2),
psychophysiological insomnia (PI) is a form of insomnia that comprises heightened
arousal and learned sleep-preventing associations, which cause insomnia and poor
functioning during wakefulness. Patients with PI frequently complain of sleep-
related worries and ruminations that may reduce cognitive performance.
Hyperarousal is an important model to understand the pathophysiology of
insomnia. In terms of sleep–wake regulation neurobiology, insomnia is presumed
to result from dysregulation of arousal networks including the hypothalamus, brain
stem, and cortical areas. While psychophysiological experiments related to
hyperarousal insomnia have been actively investigated, there have been few task-
related neuroimaging studies involving the provision of sleep-related stimuli for
evaluation of hyperarousal reactions in patients with PI

19
2.7 TITLE: DISCRIMINATING PARADOXICAL AND
PSYCHOPHYSIOLOGICAL INSOMNIA BASED ON
STRUCTURAL AND FUNCTIONAL BRAIN IMAGES: A
PRELIMINARY MACHINE LEARNING STUDY
AUTHOR: MORTAZA AFSHANI YEAR 2022

Recent studies have suggested that there are various ID subtypes with different
ethology and distinct pathophysiology. Subtyping ID is an ongoing debate in sleep
medicine and has changed over different versions of the International
Classification of Sleep Disorders (ICSD). In ICSD-2, several ID subtypes were
introduced including paradoxical insomnia (PDI) and psychophysiological
insomnia (PPI). ICSD-2 defines PDI as subjective complaints of insomnia, while
polysomnography (PSG) shows near-normal sleep patterns, causing a discrepancy
between subjective and objective sleep measures. PPI, instead, is characterized by
increased arousal before or during sleep and learned-sleep preventing associations
in a routine bedroom setting. Put differently, the more a patient tries to sleep, the
more he/she gets anxious and more difficult to initiate sleep, which causes low
functional performance during wakefulness. Thereby, PPI is described by fear of
sleep and the bedroom environment. However, although ICSD-3 has emphasized
the existence of the ID subtypes, it proposed to consider it as a single category,
entitled chronic ID, as there are no customized therapeutic approaches for each
subtype. This preliminary study provides evidence that structural and functional
brain measures can help to distinguish two common subtypes of ID from each other
and from healthy subjects. Moreover, we observed that the multimodal brain
measure is a bit better than the unimodal brain measure to separate ID subtypes

20
 CHAPTER 3

SYSTEM ANALYSIS

3.1 EXISTING SYSTEM

Insomnia is a common sleep disorder that can lead to difficulty falling

asleep, staying asleep, or waking up too early. There are various existing systems
for insomnia sleep stage prediction that can provide valuable information about
sleep patterns. Polysomnography (PSG) is considered the gold standard for sleep
stage classification and involves monitoring various physiological signals to
identify different stages of sleep. There are also various wearable devices available
in the market that claim to track sleep stages, and machine learning models can be
used to predict sleep stages based on various physiological signals. Support Vector
Machine (SVM) is another machine learning algorithm that can be used for sleep
stage classification. SVM is a supervised learning algorithm that can classify data
into different categories by finding the best separating hyperplane. In sleep stage
classification, SVMs are trained using features extracted from PSG recordings. The
feature extraction process involves analysing the PSG recordings to extract
relevant information about the different sleep stages. The extracted features are
then used to train the SVM, which can then classify new epochs of PSG recordings
into different sleep stages. Some common features used for sleep stage
classification using SVMs include spectral features, such as power in different
frequency bands, and statistical features, such as the mean and variance of the EEG
signal. SVMs can also be combined with other machine learning algorithms, such
as k-nearest neighbours (k-NN) or decision trees, to improve classification
performance.

21
3.1.1 DISADVANTAGES

• Computational Requirements: Deep learning algorithms require significant

computational resources and time to train, which can be a challenge for
healthcare systems with limited resources.
• Time complexity is high
• Does not support large datasets
• Accuracy is less

3.2 PROPOSED METHOD

Sleep is the brain's primary function and plays a fundamental role in

individual performance, learning ability, and physical movement. One of the
essential physiological processes of humans is sleep vital for physical and
cognitive well-being and resurgence. Sleep is a reversible state in which the eyes
are closed, and several nerve centres are disabled. Sleep creates partial or unique
or full anaesthesia for the individual, in which case the brain becomes a less
complicated network. The conventional solution is to detect the artefacts and
denoise the signal by removing corresponding epochs from the sleep signal.
However, this way, the EEG signal will be manipulated and may lose important
information. One of the motivations of this thesis is to develop and improve noise
cancelation method that does not manipulate the signal and protect its originality.
The proposed system for sleep stage classification using CNNs from an EDF and
CSV file involves several steps. The first step is data pre-processing, where the raw
data is filtered, resampled, and features are extracted. This step helps to remove
noise and ensure that the data is consistent across all recordings. The next step is
data splitting, where the data is split into training, validation, and testing sets. The
training set is used to train the CNN model, the validation set is used to tune the
hyperparameters and prevent overfitting, and the testing set is used to evaluate the

22
performance of the model. Once the data is pre-processed and split, the next step
is to design the CNN architecture. The architecture should be designed to learn
relevant features from the pre-processed data and classify the sleep stages
accurately. A common approach is to use 1D or 2D CNNs with multiple
convolutional layers and pooling layers. The output of the final layer is fed to a
fully connected layer and softmax activation to predict the sleep stage. And also
predict the sleeping disorder with new datasets, then provide the precaution details .

3.3 CONVOLUTIONAL NEURAL NETWORK

Convolutional Neural Network (CNN) A CNN is a multilayer perceptron

designed specifically to recognize two dimensional shapes with a high degree of
invariance to translation, scaling, skewing, and other forms of distortion. Learning
section of this classifier is done in supervised method which includes the following
structure:

• Feature extraction;
• Feature mapping;

The weights in all layers of a CNN are learned through training. Also, the network
learns to extract its own features automatically

Fig 3.3: Convolutional neural network framework

23
This diagram shows the multiple layers such as input layer, four hidden layers, and
an output layer. This network is designed to perform image processing. The input
layer consists of 28×28 sensor neurones, receives the images of different characters
that have been approximately centred and normalized in size. We can use the CNN
algorithm framework to classify multiple sleeping stages.

3.3.1 APPLICATIONS

Sleep stage classification has several applications in the field of sleep medicine and
research, including:

• Sleep disorder diagnosis: Accurate sleep stage classification can help

diagnose sleep disorders, such as sleep apnea, insomnia, and narcolepsy,
which can have a significant impact on a person's health and quality of life.
• Sleep quality assessment: Sleep stage classification can be used to assess the
quality of a person's sleep, which can be useful in evaluating the
effectiveness of treatments for sleep disorders or other conditions that can
affect sleep.
• Sleep tracking and monitoring: Sleep stage classification can be used to track
and monitor a person's sleep patterns over time, which can provide insights
into sleep hygiene, lifestyle factors, and other factors that can affect sleep.
• Athletic performance optimization: Sleep stage classification can be used to
optimize athletic performance by identifying the best times for training and
recovery based on an individual's sleep patterns.

Personalized medicine: Sleep stage classification can be used to develop

personalized treatment plans for sleep disorders, based on an individual's unique
sleep patterns and physiology.

24
3.3.2 ADVANTAGES

• Reduce the time and computational steps

• False positive rate is low
• Handle the large number datasets
• Layer based classification

25
 CHAPTER 4
SYSTEM IMPLEMENTATION

4.1 DATASETS

Sleep is the brain's primary function and plays a fundamental role in

individual performance, learning ability, and physical movement. One of the
essential physiological processes of humans is sleep vital for physical and
cognitive well-being and resurgence. Sleep is a reversible state in which the eyes
are closed, and several nerve centres are disabled. Sleep creates partial or unique
or full anaesthesia for the individual, in which case the brain becomes a less
complicated network. The gold standard of sleep analysis, Polysomnography
(PSG), includes measurements of several body functions, including brain activity
and heart rhythm. The sleep stages are then manually classified. These factors
cause the method to have high accuracy, but also makes it costly. The fact that the
procedure is usually conducted in a sleep laboratory or hospital can have a negative
impact on the sleep quality of the subject, in addition to the discomfort of wearing
the equipment. This is a weakness because the goal is usually to analyse the normal
sleep patterns of the subject. Another way to analyse sleep is actigraphy. An
actigraph is a body-worn sensor, consisting of a three-dimensional accelerometer
and possibly other sensors. The body movements data can be analysed in a variety
of ways. Because of the inexpensive equipment and low intrusiveness of the
method, it is preferable in some situations, for instance when the subjects are
children, when data collection for several days is necessary, or when a large group
of people is participating in a study. In this research work, an effective and robust
method is applied to classify the sleep stages automatically based on the selected
optimal set of features with an ensemble learning stacking model. The main
purpose of this study is to analyze the effectiveness of selected features with a
combination of ensemble learning for multi-class sleep stages classification

26
problems. The proposed approach considers two different categories of sleep
recordings which include subjects’ effects with different types of sleep related
disorders and the other category is subject having complete healthy control. The
proposed work is divided into two phases, in the first phase identifying the suitable
features from the extracted feature vector through obtaining different selection
algorithms. In the second phase, an ensemble learning stacking model is considered
for the classification of the sleep stages

Today, PSG is done to identify various disorders based on the analysis of

sleep stages, the main component of which is the measurement of brain activity
with EEG signal. Sleep disorders include several disorders associated with various
symptoms such as insomnia, respiratory disorders, behavioral and motor-related
sleep disorders that significantly affect EEG signals. By classifying the different
stages of sleep, the correct diagnosis of sleep-related disorders can be achieved.
After recording the EEG signal, feature extraction, and analysis of the signal
recorded in a specified range, a classification algorithm to identify the sleep stage
is used. However, improving the classification accuracy and reducing complexity
are two main challenges in the classification of sleep stages. Fig 1 shows the
different stages of sleeping cycles.

The first step is actual Deep Sleep

sleep Decrease of eye
Decrease of body
movement
temperature and blood
pressure

Deepest Sleep
Reduce muscle tension
Walk and talk in sleep

Period of transition Many dreams

between sleep and Increase of eye
alertness movement
Light sleep Increase of heart rate,
breathing and blood
27 pressure
Figure 1: Different stages of Sleep

The EEG datasets created in the Python framework and connected to

different stages of sleep are plotted in this investigation. The various classification
methods are evaluated and contrasted using accuracy as the primary criterion.
Repeated random sub-sampling validation is used to assess the effectiveness of the
sleep stage classification.

4.2 HARDWARE REQUIREMENTS

• Processor : Intel core processor 2.6.0 GHZ

• RAM : 4 GB
• Hard disk : 160 GB
• Compact Disk : 650 Mb
• Keyboard : Standard keyboard
• Monitor : 15 inch color monitor

4.3 SOFTWARE REQUIREMENTS

• Operating System : Windows OS

• Front-End : PYTHON
• IDE : PYCHARM
• Libraries : Tensorflow, KERAS

4.4 MODULES
4.4.1 DATASETS ACQUISITION

In this module we can input the CSV file about EEG data and also EDF file.
The Physionet dataset is a dataset of EEG recordings for multiple persons
PhysioNet is a public repository of biomedical signals and time series data, which
can be used for research and development in the field of sleep stage prediction. It

28
contains large datasets of sleep and physiological data, including EEG, ECG, and
respiration signals, which can be used to train and evaluate deep learning
algorithms for sleep stage prediction. In this module, we can input the CSV file and
EDF file. CSV file contains the frequency values. EDF (European Data Format) is
a common file format for storing physiological signals, including those recorded
during sleep. EDF files typically contain several signals, including EEG, EOG,
EMG, and ECG. These signals are recorded using electrodes placed on the scalp,
face, and body, and provide information about the electrical activity of the brain
and muscles during sleep. To use an EDF file for sleeping stage prediction, the first
step is to load the file into a software program or Python library that can read EDF
files. Examples of such libraries include MNE-Python and pyEDFlib. Once the
EDF file is loaded, the next step is to extract features from the signals that are
relevant to sleep stage prediction. Common features include the power spectrum,
the amplitude of specific frequency bands, and the correlation between different
signals.

Input: CSV file Irrelevant data

removal

Missing data Output: Structured

removal datasets

4.4.2 PREPROCESSING
Preprocessing is an important step in the analysis of EEG data, as it helps to
remove artifacts and improve the quality of the data. In the pre-processing stage,
continuous EEG recordings are firstly segmented without overlapping by a sliding
time window. In this module, perform preprocessing steps to eliminate the

29
irrelevant and missing datasets from uploaded CSV file. The first step is to load
the EDF file into memory using a library such as MNE-Python or pyEDFlib. The
CSV file containing the sleep stage annotations can also be loaded into a Pandas
DataFrame. Sleep signals are often contaminated by noise and artifacts. Signal
filtering can be used to remove these unwanted components and improve the
quality of the signal. Common filtering techniques include bandpass filtering,
notch filtering, and high-pass filtering.

Input: CSV file Irrelevant data

removal

Missing data Output: Structured

removal datasets

4.4.3 FEATURES EXTRACTION:

Feature extraction is the process of transforming raw data into a set of
features that can be used to represent the data in a more meaningful and useful way.
Relevant features need to be extracted from the signal to represent the underlying
physiological processes. Common features for sleep stage classification include
power spectral density, amplitude of specific frequency bands, and correlation
between different signals. In the context of a Convolutional Neural Network
(CNN) for EEG data, feature extraction involves transforming the raw EEG signals
into a set of features that can be used as input to the CNN. In this module, we can
calculate the mean and standard deviation for uploaded CSV file. Normalizing the
data is important to ensure that all features have the same scale and range. This
helps to improve the performance of the machine learning algorithm. In CSV file,
train the Y_values with stage and X_train values contain the frequency values.

30
Based on features model file can be created using CNN algorithm for future
verification.

CNN model

Input: Structured Features extraction

datasets Output: Model file
and confusion matrix

Input: Structured
datasets Model
construction
Output: Model file

4.4.4 CLASSIFICATION:
The features are divided into training, validation, and test sets, and are
organized into a format that can be used as input to the CNN. A CNN is designed
and configured, including the choice of architecture, number of layers, and
activation functions. The CNN is trained on the training set, using a loss function
and an optimization algorithm. The model is validated on the validation set to
prevent overfitting and monitor the performance of the model. The trained CNN is
evaluated on the test set to assess its performance in classifying the sleep stages. A
CSV (Comma-Separated Values) file can be used for sleep stage classification in
EEG datasets by providing annotations for each epoch of data. An epoch is a fixed-
length segment of EEG data, typically 30 seconds in length. The annotations in the
CSV file indicate the sleep stage that each epoch belongs to, as determined by a
sleep expert. To classify EEG datasets using a CSV file, the first step is to load the
EEG data and the corresponding CSV file into memory. The EEG data can be
loaded using a library such as MNE-Python or pyEDFlib, while the CSV file can
be loaded into a Pandas DataFrame. Once the data is loaded, the next step is to

31
preprocess the EEG data to extract features that are relevant for sleep stage
classification. This can involve techniques such as filtering the data, resampling
the data, and extracting features such as power spectral density, amplitude of
specific frequency bands, and correlation between different EEG channels. A
common approach for EEG classification is to use a convolutional neural network
(CNN). The CNN is trained on the labeled epochs to learn the relationship between
the EEG features and the sleep stages. The performance of the CNN is evaluated
using metrics such as accuracy, precision, recall, and F1 score.

4.4.5 PREDICTION
Finally, the trained CNN can be used to classify the sleep stages in new EEG
datasets. The EEG data is divided into epochs, and the CNN predicts the sleep stage
for each epoch. This information can be used to analyze the sleep patterns of
individuals and can provide valuable insights for clinical and research purposes. In
this module, input the EEG readings in terms of numeric format. And match the
reading with CNN model file. Then classify the stages of sleeping disorders and
provide the precautions to users

Sleeping stage prediction

Input: EEG values Model deployment

Stage Output: Precaution

classification details

32
 CHAPTER 5
SYSTEM DESIGN

5.1 SYSTEM ARCHITECTURE

33
5.2 UML DIAGRAMS

5.2.1 USE CASE DIAGRAM

A use case diagram is a type of UML (Unified Modeling Language) diagram

that is used to represent the functional requirements of a system or software
application. It provides a high-level view of the system's functionality and the
actors or users who interact with it. In a use case diagram, use cases are represented
by ovals, and actors are represented by stick figures or icons. The use cases describe
the various interactions that the actors can have with the system, and the
relationships between the actors and the use cases.

Datasets Acquisiton

Preprocessing

Training_phase

Model building

Testing_phase

Input CSV data

Prediction

34
5.2.2 CLASS DIAGRAM

A class diagram is a type of UML (Unified Modeling Language) diagram

used to represent the static structure of a system or software application. It shows
the classes and interfaces that make up the system, as well as the relationships and
dependencies between them.

Training phase Testing phase

+CSV file +CSV data

+EDF file +Input the CSV file()
+Upload the datasets() +Classification()
+Preprocessing() +Prediction details()
+Features extraction()
+Model building -CNN()
+Analysis report()

5.2.3 ACTIVITY DIAGRAM

An activity diagram is a type of UML (Unified Modelling Language)
diagram used to describe the flow of activities and actions in a system or process.
It depicts the steps and decisions involved in a particular activity or process and
shows the interactions between different actors, objects, or system components.
Activity diagrams are often used to model the behaviour of software systems,
business processes, or workflows. They are a visual representation of the steps
required to complete a task or achieve a specific goal, and can be used to
communicate complex processes to stakeholders.

35
 Datasets acquisition

Preprocessing

Features extraction

Classification

Sleeping stage classification

5.2.4 SEQUENCE DIAGRAM

A sequence diagram is a type of UML (Unified Modeling Language)

diagram that is used to visualize the interactions between objects or components in
a system or software application. It depicts the sequence of messages exchanged
between objects and the order in which they occur. In a sequence diagram, objects
are represented by rectangles, and the messages exchanged between objects are
represented by arrows. The objects are listed vertically on the left-hand side of the
diagram, and the messages are listed horizontally across the top of the diagram.

36
 Datasets Acquisition Preprocessing Features extraction Classification
Prediction

1 : CSV or EEG datasets()

2 : Irrelevant and missing data removal()

3 : Time and frequency domain features()

4 : CNN based model build()

5 : Sleeping stage classification()

6 : Performance evaluation()

5.2.5 COLLABORATION DIAGRAM

A collaboration diagram, also known as a communication diagram, is an

illustration of the relationships and interactions among software objects in the
Unified Modeling Language (UML). Developers can use these diagrams to portray
the dynamic behavior of a particular use case and define the role of each object.

37
38
 CHAPTER 6

SYSTEM TESTING

The purpose of testing is to discover errors. Testing is the process of trying

to discover every conceivable fault or weakness in a work product. It provides a
way to check the functionality of components, sub-assemblies, assemblies and/or
a finished product. It is the process of exercising software with the intent of
ensuring that the Software system meets its requirements and user expectations and
does not fail in an unacceptable manner. There are various types of tests. Each test
type addresses a specific testing requirement.

6.1 SOFTWARE TESTING STRATEGIES

Testing involves

➢ Unit Testing
➢ Functional Testing
➢ Acceptance Testing

6.1.1 UNIT TESTING

Unit testing involves the design of test cases that validate that the internal
program logic is functioning properly, and that program inputs produce valid
outputs. All decision branches and internal code flow should be validated. It is the
testing of individual software units of the application. It done after the completion
of an individual unit before integration. This is a structural testing, that relies on
knowledge of its construction and is invasive. Unit tests perform basic tests at
component level and test a specific business process, application, and/or system
configuration. Unit tests ensure that each unique path of a business process
performs accurately to the documented specifications and contains clearly defined
inputs and expected results.

39
6.1.2 FUNCTIONAL TESTING

Functional tests provide systematic demonstrations that functions tested are

available as specified by the business and technical requirements, system
documentation, and user manuals.

Functional testing is centered on the following items:

Valid Input : identified classes of valid input must be accepted.

Invalid Input : identified classes of invalid input must be rejected.

Functions : identified functions must be exercised.

Output : identified classes of application outputs must be exercised.

Systems/Procedures: interfacing systems or procedures must be invoked.

Organization and preparation of functional tests is focused on requirements,

key functions, or special test cases. In addition, systematic coverage pertaining to
identify Business process flows; data fields, predefined processes, and successive
processes must be considered for testing. Before functional testing is complete,
additional tests are identified and the effective value of current tests is determined.

6.1.3 ACCEPTANCE TESTING

User Acceptance Testing is a critical phase of any project and requires

significant participation by the end user. It also ensures that the system meets the
functional requirements. Acceptance testing is a type of software testing that aims
to determine whether a system or application meets the specified requirements and
is acceptable for delivery to the end-user. The purpose of acceptance testing is to
evaluate the system's compliance with the business requirements and to ensure that
it meets the expectations of the stakeholders.

40
 CHAPTER 7

RESULTS AND ANALYSIS

In this study we can input the EEG datasets related to sleeping stages and
developed in Python framework. Accuracy as main criterion is considered for
evaluating and comparing the different classification methods. The performance of
the sleep stage classification is evaluated using repeated random sub-sampling
validation. To measure the classification accuracy, the overall accuracy value is
calculated as follows

Number of true detections

ACCURACY =
Total number of epochs

ALGORITHMS ACCURACY (%)

SUPPORT VECTOR MACHINE 60
DEEP NEURAL NETWORK 65
CONVOLUTIONAL NEURAL NETWORK 70

ACCURACY (%)
72
70
68
66
64
62
60
58
56
54
SUPPORT VECTOR DEEP NEURAL NETWORK CONVOLUTIONAL NEURAL
MACHINE NETWORK

Fig 3: Accuracy

From the diagram, convolutional neural network provides the improved

accuracy than the existing frameworks

41
 CHAPTER 8

SNAPSHOTS

DATASETS

Number of records of Non sleep 9200 VS sleep 2300

count mean std min 25% 50% 75% max

X1 9200.0 -8.992609 70.455286 -566.0 -44.0 -7.0 26.0 1726.0

X2 9200.0 -8.877174 70.560110 -609.0 -44.0 -7.0 27.0 1713.0

X3 9200.0 -8.910435 70.372582 -594.0 -45.0 -7.0 28.0 1697.0

X4 9200.0 -8.969783 70.030409 -549.0 -45.0 -8.0 27.0 1612.0

X5 9200.0 -9.085326 69.377958 -603.0 -45.0 -8.0 27.0 1437.0

... ... ... ... ... ... ... ... ...

42
X175 9200.0 -9.848587 69.550894 -570.0 -45.0 -9.0 27.0 1958.0

X176 9200.0 -9.620435 70.353607 -594.0 -46.0 -8.0 27.0 2047.0

X177 9200.0 -9.395435 70.934300 -563.0 -45.0 -9.0 27.0 2047.0

X178 9200.0 -9.240435 71.185850 -559.0 -45.0 -8.0 27.0 1915.0

y 9200.0 0.000000 0.000000 0.0 0.0 0.0 0.0 0.0

[179 rows x 8 columns]

count mean std min 25% 50% 75% max

X1 2300.0 -21.936522 342.361939 -1839.0 -193.25 -16.0 159.00 1314.0

X2 2300.0 -19.049130 343.398782 -1838.0 -191.25 -18.0 168.25 1356.0

X3 2300.0 -15.293913 337.489643 -1835.0 -187.00 -12.5 169.25 1274.0

X4 2300.0 -9.836087 332.354833 -1845.0 -184.00 -6.0 166.25 1226.0

X5 2300.0 -3.707391 332.211163 -1791.0 -174.25 -12.0 170.00 1518.0

... ... ... ... ... ... ... ... ...

X175 2300.0 -25.830870 339.650467 -1863.0 -195.00 -14.5 153.25 1205.0

X176 2300.0 -25.043913 335.747017 -1781.0 -192.00 -18.0 150.00 1371.0

X177 2300.0 -24.548261 335.244512 -1727.0 -190.25 -21.5 151.25 1445.0

X178 2300.0 -24.016522 339.819309 -1829.0 -189.00 -23.0 157.25 1380.0

y 2300.0 1.000000 0.000000 1.0 1.00 1.0 1.00 1.0

43
[179 rows x 8 columns]

Totall Mean VALUE for sleep : 290.129360958884

Totall Std VALUE for sleep : 53.56315864740058

Totall Mean VALUE for NON sleep : 1260.098927262616

Totall Std VALUE for NON sleep : 15.561044289100993

Number of records of Non Sleep 18400 VS Sleep 18400

Normalized Totall Mean VALUE for Sleep: 1150.0132102785494

Normalized Totall Std VALUE for Sleep: 0.02256476904844563

Normalized Totall Mean VALUE for NOT Sleep: 1150.0065271687797

Normalized Totall Std VALUE for NOT Sleep: 0.002007429551034863

44
45
46
47
Model: "sequential"
_________________________________________________________________
Layer (type) Output Shape Param #
==========================================================
=======
conv1d (Conv1D) (None, 176, 256) 1024
dropout (Dropout) (None, 176, 256) 0
conv1d_1 (Conv1D) (None, 174, 128) 98432
dropout_1 (Dropout) (None, 174, 128) 0

conv1d_2 (Conv1D) (None, 172, 128) 49280

dropout_2 (Dropout) (None, 172, 128) 0
conv1d_3 (Conv1D) (None, 170, 64) 24640
dropout_3 (Dropout) (None, 170, 64) 0
conv1d_4 (Conv1D) (None, 168, 32) 6176
dropout_4 (Dropout) (None, 168, 32) 0
max_pooling1d (MaxPooling1D (None, 84, 32) 0
)
flatten (Flatten) (None, 2688) 0
dense (Dense) (None, 5) 13445
==========================================================
=======
Total params: 192,997
Trainable params: 192,997
Non-trainable params: 0

precision recall f1-score support

0 0.99 0.94 0.96 465

48
 1 0.63 0.69 0.66 459
2 0.74 0.40 0.52 450
3 0.89 0.79 0.84 457
4 0.71 0.88 0.79 469
micro avg 0.78 0.74 0.76 2300
macro avg 0.79 0.74 0.75 2300
weighted avg 0.79 0.74 0.75 2300
samples avg 0.74 0.74 0.74 2300
accuracy: 76.43%

49
50
51
Accuracy score: 0.6699909338168631
[155 0 0 2 0]
[ 96 5 4 1 3]
[ 78 12 443 21 8]
[ 0 0 5 100 0]
[ 84 27 23 0 36]]
precision recall f1-score support
Sleep stage W 0.38 0.99 0.54 157
Sleep stage 1 0.11 0.05 0.07 109
Sleep stage 2 0.93 0.79 0.85 562
Sleep stage 3/4 0.81 0.95 0.87 105
Sleep stage R 0.77 0.21 0.33 170

accuracy 0.67 1103

macro avg 0.60 0.60 0.53 1103
weighted avg 0.73 0.67 0.65 1103

52
 CHAPTER 9

APPENDIX

from tkinter import

import os
from tkinter import filedialog
from tkinter import messagebox
import pandas as pd
import tensorflow as tf
import numpy as np
from keras.models import Sequential
from keras.utils import np_utils
from keras.layers import Dense, Activation, Conv1D, MaxPooling1D, Dropout,
Flatten
from sklearn.utils import shuffle
from sklearn.preprocessing import normalize, StandardScaler
from sklearn.model_selection import train_test_split
import matplotlib.pyplot as plt
def endprogram():
print("\nProgram terminated!")
sys.exit()
def upload():
# import cufflinks
# cufflinks.go_offline()
# cufflinks.set_config_file(world_readable=True, theme='pearl')
import plotly
import plotly.figure_factory as ff

53
 from plotly import tools
from plotly.offline import init_notebook_mode, iplot
import imblearn
init_notebook_mode(connected=True)
pd.set_option('display.max_columns', 100)
data = pd.read_csv('Data/Sleepdataset.csv')
data.head()
data.tail()
data.shape
data['y'].value_counts()
data.y.hist();
plt.figure(figsize=(50, 4))
plt.subplot(131)
[plt.plot(data.values[i][1:-1]) for i in range(23)];
dic = {5: 0, 4: 0, 3: 0, 2: 0, 1: 1}
data['y'] = data['y'].map(dic)
print(data['y'].value_counts())
data.head()
data = data.drop('Unnamed', axis=1)
data = shuffle(data)
data.describe()
data.info()
print('Number of records of Non sleep {0} VS sleep
{1}'.format(len(data[data['y'] == 0]),
len(data[data['y'] == 1])))
data[data['y'] == 0].describe().T

54
print(data[data['y'] == 0].describe().T)
data[data['y'] == 1].describe().T
print(data[data['y'] == 1].describe().T)
print('Totall Mean VALUE for sleep : {}'.format((data[data['y'] ==
1].describe().mean()).mean()))
print('Totall Std VALUE for sleep : {}'.format((data[data['y'] ==
1].describe().std()).std()))
range(2)];
[(plt.figure(figsize=(8, 4)), plt.title('Sleep'), plt.plot(data[data['y'];
def indic(data):
max = np.max(data, axis=1)
min = np.min(data, axis=1)
return max, min
x1, y1 = indic(Not_Schizophrenia)
x2, y2 = indic(Schizophrenia)
fig = plt.figure(figsize=(14, 6))
ax1 = fig.add_subplot(111)
ax1.scatter(x1, y1, s=10, c='b', label='Not Sleep')
ax1.scatter(x2, y2, s=10, c='r', label='Sleep')
plt.legend(loc='lower left');
plt.show()
# Just Epileptic
x, y = indic(data[data['y'] == 1].iloc[:, range(0, len(data.columns) - 1)]
plt.figure(figsize=(14, 4))
plt.title('Sleep')
plt.scatter(x, y, c='r');
plt.show()

55
x, y = indic(data[data['y'] == 0].iloc[:, range(0, len(data.columns) - 1)]
plt.figure(figsize=(14, 4))
plt.title('NOT Sleep')
plt.scatter(x, y);
plt.show()
# define oversampling strategy
# fit and apply the transform
X, y = oversample.fit_resample(data.drop('y', axis=1), data['y'])
X.shape, y.shape
print('Number of records of Non Sleep {0} VS Sleep {1}'.format(len(y == True),
normalized_df = pd.DataFrame(normalize(X))
normalized_df
normalized_df['y'] = y
print('Normalized Totall Mean VALUE for Sleep: {}'.format(
(normalized_df[normalized_df['y'] == 1].describe().mean()).mean()))
print( 'Normalized Totall Std VALUE for Sleep: {}'.format(
(normalized_df[normalized_df['y'] == 1].describe().std()).std()))
print('Normalized Totall Mean VALUE for NOT Sleep: {}'.format(
(normalized_df[normalized_df['y'] == 0].describe().mean()).mean()))
print('Normalized Totall Std VALUE for NOT Sleep: {}'.format(
(normalized_df[normalized_df['y'] == 0].describe().std()).std()))
def training():
data = pd.read_csv('Data/Sleepdataset.csv')
data.head()
from sklearn.preprocessing import LabelBinarizer
X = np.asarray(data.values)

56
X = np.asarray(X[:, 1:-1])
X = X.astype(float)
minm = X.min()
maxm = X.max()
X_norm = (X - minm) / (maxm - minm)
print(X.shape)
from sklearn.model_selection import train_test_split
mlb = LabelBinarizer()
y = np.asarray(data['y'])
# y=np.array.astype('float')
# Y=mlb.fit_transform(y)
Y = y.astype(float)
# X=X.reshape(1200,178,1)
# print(X.shape)
Y=Y-1
from tensorflow.python.keras.utils.np_utils import to_categorical
Y = to_categorical(Y)
X_train, X_test, Y_train, Y_test = train_test_split(X, Y, test_size=0.20,)
model = Sequential()
model.add(Conv1D(filters=256, kernel_size=3, activation='relu')
model.add(Dropout(0.3))
model.add(Conv1D(filters=128, kernel_size=3, activation='relu'))
model.add(Dropout(0.3))
model.add(Conv1D(filters=128, kernel_size=3, activation='relu'))
model.add(Dropout(0.3))
model.add(Conv1D(filters=64, kernel_size=3, activation='relu'))

57
model.add(Dropout(0.3))
model.add(Conv1D(filters=32, kernel_size=3, activation='relu'))
model.add(Dropout(0.3))
model.add(MaxPooling1D(pool_size=2))
model.add(Flatten())
# model.add(Dense(32, activation='relu'))
model.add(Dense(5, activation='softmax'))
# model.compile(loss='categorical_crossentropy', optimizer='adam')
model.summary()
model.compile(optimizer='adam', loss='binary_crossentropy')
y_ints = [y.argmax() for y in Y_train]
from sklearn.utils import class_weight
class_weights = class_weight.compute_class_weight('balanced',np.unique(y_ints)
history2 = model.fit(X_train, Y_train, 100, 150, verbose=1)
# class_weight=class_weight_dict)
model.save('cnn.hdf5')
acc = history2.history['accuracy']
loss = history2.history['loss']
epochs = range(1, len(acc) + 1)
plt.plot(epochs, acc, 'b', label=' accurarcy')
plt.title('accurarcy')
plt.legend()
plt.show()
def eeeg():
import numpy as np
import matplotlib.pyplot as plt

58
import mne
from mne.datasets.sleep_physionet.age import fetch_data
from sklearn.ensemble import RandomForestClassifier
from sklearn.metrics import accuracy_score
from sklearn.metrics import confusion_matrix
from sklearn.metrics import classification_report
from sklearn.pipeline import make_pipeline
from sklearn.preprocessing import FunctionTransformer
#import pandas as pd
#sleep_data = pd.read_csv('sleep.csv')
#sleep_data.info()
import numpy as np
import matplotlib.pyplot as plt
import mne
from mne.datasets.sleep_physionet.age import fetch_data
# from mne.time_frequency import
ALICE, BOB = 0, 1
[alice_files, bob_files] = fetch_data(subjects=[ALICE, BOB], recording=[1])
mapping = {'EOG horizontal': 'eog',
'Resp oro-nasal': 'misc',
'EMG submental': 'misc',
'Temp rectal': 'misc',
'Event marker': 'misc'}
raw_train = mne.io.read_raw_edf(alice_files[0])
annot_train = mne.read_annotations(alice_files[1])
raw_train.set_annotations(annot_train, emit_warning=False)

59
raw_train.set_channel_types(mapping)
# plot events
mne.viz.plot_events(events_train, event_id=event_id,
sfreq=raw_train.info['sfreq'])
# keep the color-code for further plotting
stage_colors = plt.rcParams['axes.prop_cycle'].by_key()['color']
plt.show()
tmax = 30. - 1. / raw_train.info['sfreq'] # tmax in included
epochs_train = mne.Epochs(raw=raw_train, events=events_train,
event_id=event_id, tmin=0., tmax=tmax, baseline=None)
print(epochs_train)
raw_test = mne.io.read_raw_edf(bob_files[0], stim_channel='Event marker',
infer_types=True)
annot_test = mne.read_annotations(bob_files[1])
annot_test.crop(annot_test[1]['onset'] - 30 * 60,
annot_test[-2]['onset'] + 30 * 60)
raw_test.set_annotations(annot_test, emit_warning=False)
events_test, _ = mne.events_from_annotations(
raw_test, event_id=annotation_desc_2_event_id, chunk_duration=30.)
epochs_test = mne.Epochs(raw=raw_test, events=events_test)
tmin=0., tmax=tmax, baseline=None)
print(epochs_test)
# visualize Alice vs. Bob PSD by sleep stage.
fig, (ax1, ax2) = plt.subplots(ncols=2)
# iterate over the subjects
stages = sorted(event_id.keys())

60
for ax, title, epochs in zip([ax1, ax2],
['Alice', 'Bob'],
[epochs_train, epochs_test]):
for stage, color in zip(stages, stage_colors):
spectrum = epochs[stage].compute_psd(fmin=0.1, fmax=20.)
spectrum.plot(ci=None, color=color, axes=ax,
show=False, average=True, spatial_colors=False)
ax.set(title=title, xlabel='Frequency (Hz)')
ax1.set(ylabel='µV²/Hz (dB)')
ax2.legend(ax2.lines[2::3], stages)
plt.show()
# specific frequency bands
FREQ_BANDS = {"delta": [0.5, 4.5],
"theta": [4.5, 8.5],
"alpha": [8.5, 11.5],
"sigma": [11.5, 15.5],
"beta": [15.5, 30]}
spectrum = epochs.compute_psd(picks='eeg', fmin=0.5, fmax=30.)
psds, freqs = spectrum.get_data(return_freqs=True)
# Normalize the PSDs
psds /= np.sum(psds, axis=-1, keepdims=True)
X = []
for fmin, fmax in FREQ_BANDS.values():
psds_band = psds[:, :, (freqs >= fmin) & (freqs < fmax)].mean(axis=-1)
X.append(psds_band.reshape(len(psds), -1))
return np.concatenate(X, axis=1)

61
 pipe = make_pipeline(FunctionTransformer(eeg_power_band, validate=False),
RandomForestClassifier(n_estimators=100, random_state=42))
# Train
y_train = epochs_train.events[:, 2]
pipe.fit(epochs_train, y_train)
# Test
y_pred = pipe.predict(epochs_test)
# Assess the results
y_test = epochs_test.events[:, 2]
acc = accuracy_score(y_test, y_pred)
print("Accuracy score: {}".format(acc))
print(confusion_matrix(y_test, y_pred))
print(classification_report(y_test, y_pred, target_names=event_id.keys()))
def predict1():
import tensorflow as tf
import numpy as np
from tkinter import messagebox
model = tf.keras.models.load_model('Model/cnn.hdf5')
data = np.array([-9,-65,-98,-102,-78,-48,-16,0,-21,-59,-90,-103,-84,-43,-
9,3,21,-27,-96,-103,-75,-29,14,55,78,73,28,-13,-43,-68,-78,-75,-55,-41])
pred = model.predict(data, verbose=0, batch_size=200)
ind = np.argmax(pred)
print(ind)
out = ""
med =''
if ind == 0:
out='Normal'
62
elif ind == 1:
out = 'Stage1'
med ="Benzodiazepines are rapid eye movement (REM) sleep–suppressant,"
elif ind == 2:
out = 'Stage2'
med = "Benzodiazepines are rapid eye movement (REM) sleep–suppressant ,"
elif ind == 3:
out = 'Stage3'
med = "Benzodiazepines are rapid eye movement (REM) sleep–suppressant,"
elif ind == 4:
out = 'Stage4'
med = "Benzodiazepines are rapid eye movement (REM) sleep–suppressant,"
print(out)
messagebox.showinfo("Prediction", "Prediction Result:"+ out)
messagebox.showinfo("Remedy", med)
def Adminlog():
global login_screen
login_screen = Toplevel(main_screen)
login_screen.title("Predict")
login_screen.geometry("600x280")
login_screen.title("Login Form")
global username_verify
global password_verify
username_verify = StringVar()
password_verify = StringVar()
global username_login_entry

63
global password_login_entry
label_0 = Label(login_screen, text="Predict form", width=20, font=("bold",20))
label_0.place(x=90, y=53)
label_1 = Label(login_screen, text="Enter Value", width=20, font=("bold", 10))
label_1.place(x=80, y=130)
username_login_entry = Entry(login_screen, textvariable=username_verify)
username_login_entry.place(x=240, y=130)
bluebutton = Button(login_screen, text="Predict", fg="blue", font=('helvetica')
bluebutton.place(x=220, y=210)
def main_account_screen():
global main_screen
main_screen = Tk()
width = 600
height = 600
screen_width = main_screen.winfo_screenwidth()
screen_height = main_screen.winfo_screenheight()
x = (screen_width / 2) - (width / 2)
y = (screen_height / 2) - (height / 2)
main_screen.geometry("%dx%d+%d+%d" % (width, height, x, y))
main_screen.resizable(0, 0)
# main_screen.geometry("300x250")
main_screen.title("Sleep Abnormal Prediction")
Label(text="Sleep Abnormal Prediction", bg="Blue", width="300”,height="5")
Button(text="Upload Dataset", font=(
'Verdana', 15), height="2", width="30", command=upload,
Label(text="").pack()

64
Button(text="Training", font=(
'Verdana', 15), height="2", width="30", command=training,
Label(text="").pack()
Button(text="Predict", font=(
'Verdana', 15), height="2", width="30", command=Adminlog,
Label(text="").pack()
Button(text="EEG Based Evaluation", font=('Verdana', 15), height="2",
width="30",command=eeeg).pack(side=TOP)
Label(text="").pack()
main_screen.mainloop()
main_account_screen()

65
 CHAPTER 10

CONCLUSION AND FUTURE WORK

9.1 CONCLUSION

In conclusion, using a CSV file for sleep stage classification in EEG datasets
can provide valuable information about an individual's sleep patterns. By loading
the EEG data and CSV file into memory and pre-processing the data to extract
relevant features, a labelled dataset of epochs can be created that can be used to
train and evaluate machine learning algorithms. A CNN is a common machine
learning algorithm used for EEG classification, which can learn the relationship
between the EEG features and the sleep stages. The performance of the CNN can
be evaluated using metrics such as accuracy, precision, recall, and F1 score. The
trained CNN can then be used to predict the sleep stage in new EEG datasets by
dividing the data into epochs and classifying each epoch based on its EEG features.
Overall, CSV file classification for EEG datasets provides a powerful tool for sleep
researchers and clinicians to analyze large amounts of sleep data quickly and
accurately. It can help to identify sleep disorders, monitor treatment progress, and
provide insights into the mechanisms of sleep.

9.2 FUTURE ENHANCEMENT

While sleeping stage classification using a CSV file and EEG datasets has
shown promising results, there is still room for future work to improve the accuracy
and efficiency of the classification. One area for future work is to explore more
advanced machine learning techniques beyond CNNs. For example, recurrent
neural networks (RNNs) could be used to incorporate temporal information in the
classification process, as sleep stages are known to have temporal dependencies.

66
 CHAPTER 11

REFERENCES

[1] Malik, Asra, et al. "Detection of Insomnia using Electrocardiography and

Electromyography." 2021 International Conference on Artificial Intelligence
(ICAI). IEEE, 2021.
[2] Yang, Bufang, and Hongxing Liu. "Automatic identification of insomnia based
on single-channel EEG labelled with sleep stage annotations." IEEE Access 8
(2020): 104281-104291.
[3] Kuo, Chih-En, and Guan-Ting Chen. "A short-time insomnia detection system
based on sleep EOG with RCMSE analysis." IEEE Access 8 (2020): 69763-69773.
[4] Andresini, Giuseppina, et al. "Insomnia: Towards concept-drift robustness in
network intrusion detection." Proceedings of the 14th ACM Workshop on Artificial
Intelligence and Security. 2021.
[5] Sharma, Manish, Harsh S. Dhiman, and U. Rajendra Acharya. "Automatic
identification of insomnia using optimal antisymmetric biorthogonal wavelet filter
bank with ECG signals." Computers in Biology and Medicine 131 (2021): 104246.
[6] Lee, Mi Hyun, et al. "Multitask fMRI and machine learning approach improve
prediction of differential brain activity pattern in patients with insomnia
disorder." Scientific reports 11.1 (2021): 1-13.
[7] Afshani, Mortaza, et al. "Discriminating paradoxical and psychophysiological
insomnia based on structural and functional brain images: a preliminary machine
learning study." (2022).
[8] Islam, Md Muhaiminul, et al. "Prediction of chronic insomnia using machine
learning techniques." 2020 11th International Conference on Computing,
Communication and Networking Technologies (ICCCNT). IEEE, 2020.
[9] Ma, Xiaofen, et al. "Functional connectome fingerprint of sleep quality in
insomnia patients: Individualized out-of-sample prediction using machine
learning." NeuroImage: Clinical 28 (2020): 102439.
[10] Wen, Ze-ying, et al. "Identification of discriminative neuroimaging markers
for patients on hemodialysis with insomnia: a fractional amplitude of low

67
frequency fluctuation-based machine learning analysis." BMC Psychiatry 23.1
(2023): 1-13.
[11] Rim, Beanbonyka, et al. "Deep learning in physiological signal data: A
survey." Sensors 20.4 (2020): 969.
[12] Mostafa, Sheikh Shanawaz, et al. "A systematic review of detecting sleep
apnea using deep learning." Sensors 19.22 (2019): 4934.
[13] Yildirim, Ozal, Ulas Baran Baloglu, and U. Rajendra Acharya. "A deep
learning model for automated sleep stages classification using PSG
signals." International journal of environmental research and public health 16.4
(2019): 599.
[14] Craik, Alexander, Yongtian He, and Jose L. Contreras-Vidal. "Deep learning
for electroencephalogram (EEG) classification tasks: a review." Journal of neural
engineering 16.3 (2019): 031001.
[15] Hong, Shenda, et al. "Opportunities and challenges of deep learning methods
for electrocardiogram data: A systematic review." Computers in Biology and
Medicine 122 (2020): 103801.
[16] Lashgari, Elnaz, Dehua Liang, and Uri Maoz. "Data augmentation for deep-
learning-based electroencephalography." Journal of Neuroscience Methods 346
(2020): 108885.
[17] Zhang, Xiang, et al. "A survey on deep learning-based non-invasive brain
signals: recent advances and new frontiers." Journal of neural engineering 18.3
(2021): 031002.
[18] Zhu, Tianqi, Wei Luo, and Feng Yu. "Convolution-and attention-based neural
network for automated sleep stage classification." International Journal of
Environmental Research and Public Health 17.11 (2020): 4152.
[19] Shoeibi, Afshin, et al. "Epileptic seizures detection using deep learning
techniques: a review." International Journal of Environmental Research and Public
Health 18.11 (2021): 5780.
[20] Mekruksavanich, Sakorn, and Anuchit Jitpattanakul. "Biometric user
identification based on human activity recognition using wearable sensors: An
experiment using deep learning models" Electronics 10.3 (2021): 308.

68
[21] Zhang, Linda, et al. "Automated sleep stage scoring of the Sleep Heart Health
Study using deep neural networks." Sleep 42.11 (2019): zsz159.
[22] Hosseini, Mohammad-Parsa, Amin Hosseini, and Kiarash Ahi. "A review on
machine learning for EEG signal processing in bioengineering." IEEE reviews in
biomedical engineering 14 (2020): 204-218.
[23] Singh, Himali, Rajesh Kumar Tripathy, and Ram Bilas Pachori. "Detection of
sleep apnea from heart beat interval and ECG derived respiration signals using
sliding mode singular spectrum analysis." Digital Signal Processing 104 (2020):
102796.
[24] Kulkarni, Prathamesh M., et al. "A deep learning approach for real-time
detection of sleep spindles." Journal of neural engineering 16.3 (2019): 036004.
[25] Fukazawa, Yusuke, et al. "Smartphone-based mental state estimation: A
survey from a machine learning perspective." Journal of Information Processing 28
(2020): 16-30.
[26] da Silveira, T.L.T.; Kozakevicius, A.J.; Rodrigues, C.R. Single-channel EEG
sleep stage classification based on a streamlined set of statistical features in wavelet
domain. Med. Biol. Eng. Comput. 2017.
[27] Memar, P.; Faradji, F. A Novel Multi-Class EEG-Based Sleep Stage
Classification System. IEEE Trans. Neural Syst. Rehabil. Eng. 2018.
[28] Yulita, I.N.; Fanany, M.I.; Arymurthy, A.M. Fast convolutional method for
automatic sleep stage classification. Healthc. Inform. Res. 2018
[29] Talo, M.; Baloglu, U.B.; Yıldırım, Ö.; Acharya, U.R. Application of deep
transfer learning for automated brain abnormality classification using MR images.
Cogn. Syst. Res. 2019, 54, 176–188.
[30] Faust, O.; Hagiwara, Y.; Hong, T.J.; Lih, S.; Acharya, R. Deep learning for
healthcare applications based on physiological signals: A review. Comput.
Methods Programs Biomed. 2018.

69
 International Journal of Scientific Research in Engineering and Management (IJSREM)
Volume: 07 Issue: 03 | March - 2023 Impact Factor: 7.185 ISSN: 2582-3930

SLEEPING ABNORMALITIES DETECTION USING

DEEP LEARING TECHNIQUE
Mr. SATHEESHKUMAR MCA, M.Phil,
SANTHOSH S, ANANTHANAYAKI S, SWATHIPRIYA K

Department of Computer Science and Engineering

University College Of Engineering, Thirukkuvalai

(A constituent College Of Anna University::Chennai and Approved by AICTE, New Delhi)

ABSTRACT KEYWORDS: Artificial intelligence, Insomnia,

Machine learning, Deep learning, Sleeping
Maintaining proper health and mental stage
stability is critical for overall health and wellbeing.
Despite a good deal of research investment, sleep 1. INTRODUCTION
quality continues to be a crucial public challenge.
Nowadays, people of all age groups are affected by Sleep is the brain's primary function and
improper sleep quality. Poor sleep can lead to a plays a fundamental role in individual
variety of neurological disorders. Sleep disorders are performance, learning ability, and physical
common in all subsets of the population, movement. One of the essential physiological
independently of gender. This public health processes of humans is sleep vital for physical and
challenge greatly affects quality of life in terms of cognitive well-being and resurgence. Sleep is a
both physical and mental health. Insomnia, reversible state in which the eyes are closed, and
parasomnias, sleep-related breathing difficulties, several nerve centres are disabled. Sleep creates
hypersomnia, bruxism, narcolepsy, and circadian partial or unique or full anaesthesia for the
rhythm disorders are some common examples of individual, in which case the brain becomes a less
sleep-related disorders. Some of these disorders can complicated network.
be treated with proper analysis of early symptoms; The gold standard of sleep analysis,
in such cases, adequate sleep quality is essential for Polysomnography (PSG), includes measurements
the patient’s recovery. Artificial intelligence has of several body functions, including brain activity
several applications in sleep medicine including and heart rhythm. The sleep stages are then
sleep and respiratory event scoring in the sleep manually classified. These factors cause the
laboratory, diagnosing and managing sleep method to have high accuracy, but also makes it
disorders, and population health. While still in its costly. The fact that the procedure is usually
nascent stage, there are several challenges which conducted in a sleep laboratory or hospital can
preclude AI’s generalizability and wide-reaching have a negative impact on the sleep quality of the
clinical applications. Artificial intelligence is a subject, in addition to the discomfort of wearing
powerful tool in healthcare that may improve patient the equipment. This is a weakness because the
care, enhance diagnostic abilities, and augment the goal is usually to analyse the normal sleep
management of sleep disorders. However, there is a patterns of the subject. Another way to analyse
need to regulate and standardize existing machine sleep is actigraphy. An actigraphy is a body-worn
learning algorithms and deep learning algorithm sensor, consisting of a three-dimensional
prior to its inclusion in the sleep clinic. accelerometer and possibly other sensors.

70
 International Journal of Scientific Research in Engineering and Management (IJSREM)
Volume: 07 Issue: 03 | March - 2023 Impact Factor: 7.185 ISSN: 2582-3930

The body movements data can be

analysed in a variety of ways. Because of the
inexpensive equipment and low intrusiveness of
the method, it is preferable in some situations, for
instance when the subjects are children, when
data collection for several days is necessary, or
when a large group of people is participating in a
study. In this research work, an effective and
robust method is applied to classify the sleep
stages automatically based on the selected
optimal set of features with an ensemble learning
stacking model. The main purpose of this study is
to analyse the effectiveness of selected features
with a combination of ensemble learning for
multi-class sleep stages classification problems.
The proposed approach considers two different
categories of sleep recordings which include
subjects’ effects with different types of sleep
related disorders and the other category is subject
having complete healthy control. The proposed
work is divided into two phases, in the first phase
identifying the suitable features from the Figure 1: Different stages of Sleep
extracted feature vector through obtaining
2. RELATED WORK
different selection algorithms. In the second
phase, an ensemble learning stacking model is Malik, Asra, et al, … [1] have used both
considered for the classification of the sleep ECG and EMG signals for the detection of
stages Insomnia. Our proposed model is based on rational
classification and analysis of ECG and EMG signals
Today, PSG is done to identify various
acquired by non-invasive and minimal-cost sensors.
disorders based on the analysis of sleep stages, the The advanced method is accurate in the
main component of which is the measurement of
identification of Insomnia. In this method,
brain activity with EEG signal. Sleep disorders integrating ECG and EMG signals with derived
include several disorders associated with various features and classifiers has provided 100% accuracy
symptoms such as insomnia, respiratory but the system needs to be realized on hardware
disorders, behavioural and motor-related sleep
which needs further investigation. The Cyclic
disorders that significantly affect EEG signals. By Alternating Pattern (CAP) database from Physio net
classifying the different stages of sleep, the
is used. Pre-processing by empirical mode
correct diagnosis of sleep-related disorders can be decomposition (EMD) is carried out in this research.
achieved. After recording the EEG signal, feature
Features with high discriminative ability are
extraction, and analysis of the signal recorded in extracted from signals to classify via logistic
a specified range, a classification algorithm to
regression (LR), support vector machines (SVM),
identify the sleep stage is used. However,
K-nearest neighbour
improving the classification accuracy and
(KNN), decision tree (DT), ensemble classifier (EC),
reducing complexity are two main challenges in
and Naïve Bayes (NB) classification methods. we
the classification of sleep stages. Fig 1 shows the
have achieved the highest accuracy of 100% on both
different stages of sleeping cycles.
ECG and EMG signals from our proposed
methodology.

Yang, Bifan, et.al,…[2] proposed a

1DCNN model for automatic insomnia
identification based on single-channel EEG

71
 International Journal of Scientific Research in Engineering and Management (IJSREM)
Volume: 07 Issue: 03 | March - 2023 Impact Factor: 7.185 ISSN: 2582-3930

labelled with sleep stage annotations, and further framework to perform a time aware evaluation of
investigated the identification performance based INSOMNIA on a recently published, revised
on different sleep stages. Our experiments version of CICIDS2017 and demonstrate that
demonstrated that our 1D-CNN leveraging the 3 modern intrusion detection systems must address
sub datasets composed of REM, LSS and SWS concept drift in order to be effective. We envision
epochs, respectively, achieved higher average that future work may build on INSOMNIA in order
accuracy in comparison with baseline methods to design robust intrusion detection models that can
under both intra-patient and inter-patient be sustained over time.
paradigms. The experimental results also To evaluate INSOMNIA, we extend
indicated that amongst all the sleep stages, 1D- TESSERACT—a framework originally proposed
CNN leveraging REM and SWS epochs exhibited for performing sound time-aware evaluations of
the best insomnia identification performance in ML-based malware detectors—to the network
intra-patient paradigm, whereas no statistically intrusion domain and show that accounting for drift
significant difference was found in inter-patient is vital for effective detection. Sharma, Manish,
paradigm. And conducted experiments under et.al,[5] have used a small number of subjects to
intra-patient and inter-patient paradigms, develop the model. This 321 is one of the limitations
respectively. Our experiments demonstrated that of this work. We intend to use more data with a
our 1D-CNN leveraging 3 sub datasets composed greater number of subjects to validate our developed
of REM, LSS and SWS epochs, respectively, system in the future. We are also planning to explore
achieved higher average accuracies in the possibility of developing a deep learning model
comparison with baseline methods under both using huge database. Further, we also plan to extend
intra-patient and inter-patient paradigms. this work using CAP database with other
physiological signals such as multi-modal EEG and
Kuo, Chih-En, et.al,…[3] implemented a EOG. Some other disorders such as bruxism and
short-time insomnia detection system based on a narcolepsy, nocturnal frontal lobe epilepsy (NFLE),
single-channel sleep EOG with RCMSE analysis periodic leg movement (PLM), rapid-eye movement
was proposed. First, a single-channel sleep EOG (REM) behavioural disorder and sleep disordered
was filtered with a band pass filter to remove breathing can also be detected using such automated
artifacts. Second, the RCMSE values with a scale models. Sleep disorders such as sleep movement
factor of 1 to 8 were extracted from the all-night disorders, nocturnal front lobe epilepsy, insomnia,
sleep EOG in 30-s epochs to compare the and narcolepsy are caused due to low sleep quality.
differences between the healthy and insomnia Insomnia is one such sleep disorder where a person
groups. Third, the RCMSE values from the first has difficulty in getting quality sleep. There is no
27.5 min, with scale factor of 1 to 9 were used to definitive test to identify insomnia; hence it is
compute its mean values as the input of classifier. essential to develop an automated system to identify
Finally, the support vector machine (SVM) was it accurately. A few automated methods have been
used to detect insomnia. In addition, MSE and proposed toidentifyinsomnia using either
RCMSE were applied to analyse the sleep EOG polysomnogram (PSG) or electroencephalogram
signals from subjects belonged the different (EEG) signals. To the best of our knowledge, we are
groups, and the MSE and RCMSE values with the first to automatically detect insomnia using only
different sleep stages were also compared electrocardiogram (ECG) signals without
between the healthy and insomnia groups combining them with any other physiological
signals.
Andresini, Giuseppina, et.al,…[4] outlined
a set of open challenges facing modern ML-based
intrusion detectors relating to a lack of uniformity in
the distribution of traffic data over time. To tackle Lee, Mi Hyun, et.al,…[6] investigated the
them, we propose INSOMNIA, a semi-supervised differential spatial covariance pattern of blood
approach that uses active learning to reduce latency oxygen level-dependent (BOLD) responses to
in the model updates, label estimation to reduce single-task and multitask functional magnetic
labelling overhead, and applies explainable AI to resonance imaging (fMRI) between patients with
describe how the model evolves to fit the shifting psychophysiological insomnia (PI) and healthy
distribution. We extend the TESSERACT controls (HCs), and evaluated features generated by

72
 International Journal of Scientific Research in Engineering and Management (IJSREM)
Volume: 07 Issue: 03 | March - 2023 Impact Factor: 7.185 ISSN: 2582-3930

principal component analysis (PCA) for Islam, Md Muhaiminul, et.al,…[8] shows

discrimination of PI from HC, compared to features that Logistic regression model performed best
generated from BOLD responses to single-task compared to other models. It will be very much
fMRI using machine learning methods. In 19 handy in real-life prediction for its outstanding
patients with PI and 21 HCs, the mean beta value for cross-validation score. But still, there have been
each region of interest (Rival) was calculated with some limitations in this study. For the lack of time,
three contrast images (i.e., sleep-related picture, we were unable to gather massive data. Though the
sleep-related sound, and Stroop stimuli). We accuracy of our model is great but if we could gather
performed discrimination analysis and compared more data, it could be greater. Again, this study is
with features generated from BOLD responses to done basically done for predicting chronic insomnia
single-task fMRI. We applied support vector in humans. So, feature was selected according to
machine analysis with a least absolute shrinkage and external symptoms only. If we had collected data on
selection operator to evaluate fve performance regular habits or lifestyle (smoking, drinking, level
metrics: accuracy, recall, precision, specificity, and of using radio-wave devices) then it could be
F2. Principal component features showed the best determined also that which factors cause insomnia
classification performance in all aspects of metrics and the work would be much beneficial. Our
compared to BOLD response to single-task fMRI. thoughts upon this work are not limited just in here.
Bilateral inferior frontal gyrus (orbital), right If we have enough scope in the future, we could
calcarine cortex, right lingual gyrus, left inferior complete this works also. But this kind of approach
occipital gyrus, and left inferior temporal gyrus were is not only expensive but also time-consuming.
identified as the most salient areas by feature Expensive tests and equipment are also not available
selection. in many developing countries. To bridge this gap,
we have decided to build an intelligent model based
Afshani, Mortaza, et.al,…[7] provides on a machine learning approach that is able to
evidence that structural and functional brain predict chronic insomnia. For acquiring best results
measures can help to distinguish two common 7 different machine learning classifiers were used
subtypes of ID from each other and from healthy where our Logistic regression model outperformed
subjects. Moreover, we observed that the all of them.
multimodal brain measure is a bit better than the
unimodal brain measure to separate ID subtypes. Ma, Xiaofen, et.al,…[9] present study
Insomnia disorder (ID) is a prevalent mental demonstrated the nodal functional connectivity
illness, which is associated with poor quality of strength predicted unseen individuals’ sleep quality
life, an increased rate of motor vehicle accidents, in both short-term/acute and chronic insomnia. We
depressive symptoms, emotion dysregulation, further revealed changes in the functional
and memory impairment. Several behavioural connectivity pattern during the transition from the
and neuroimaging studies suggested that various short-term/acute insomnia to chronic insomnia. The
subtypes of ID are existing. However, the study may have clinical value by informing the
neurobiological underpinnings of ID subtypes are diagnosis of sleep quality of insomniac patients, and
poorly understood. Here, we aimed to assess may provide novel insights into the neural basis
whether unimodal and/or multimodal whole- underlying the heterogeneity of insomnia. Using 29
brain neuroimaging measurements can short term/acute insomnia participants and 44
discriminate between two of the commonly chronic insomnia participants, we used whole brain
described ID subtypes (i.e., paradoxical and regional functional connectivity strength to predict
psychophysiological insomnia) and healthy unseen individuals’ Pittsburgh sleep quality index
subjects. We obtained T1-weighted images and (PSQI), applying the multivariate relevance vector
resting-state fMRI from 34 patients with ID and regression method. Evaluated using both leave-one-
48 healthy controls. The outcome measures were out and 10-fold cross validation, the pattern of
voxel-wise values of grey matter volume, cortical whole-brain regional functional connectivity
thickness, amplitude of low-frequency strength significantly predicted an unseen
fluctuation, degree centrality, and regional individual’s PSQI in both datasets.
homogeneity. Subsequently, we applied support
vector machines to classify subjects via unimodal Wen, Ze-Ying, et.al,…[10] highlights
and multimodal measures that HDWI patients had abnormal neural

73
 International Journal of Scientific Research in Engineering and Management (IJSREM)
Volume: 07 Issue: 03 | March - 2023 Impact Factor: 7.185 ISSN: 2582-3930

activities in the right MOG and right cerebellum, problems. It is proven that shallow learning
which might be potential neural markers for techniques are not adequate tools to discriminate
distinguishing HDWI patients from non- among stages (e.g. still there is accuracy
insomniacs, providing further support for the reduction in detecting S1).
pathological mechanism of HDWI. Insomnia is
one of the common problems encountered in the 4. PREPROCESSING
haemodialysis (HD) population, but the The presence of artefacts might lead to the
mechanisms remain unclear. we aimed to (1) misapprehension, low accuracy and distorted
detect the spontaneous brain activity pattern in quantitative results. Therefore, a pre-processing step
HD patients with insomnia (HDWI) by using is necessary to cancel artefacts and remove cropped
fractional amplitude of low frequency fluctuation epochs to magnify informative components of raw
(fALFF) method and (2) further identify brain EEG, EOG and EMG signals prior to any further
regions showing altered fALFF as neural markers analysis. Considering the fact that nowadays
to discriminate HDWI patients from those on portable devices for patient monitoring and
haemodialysis but without insomnia (Hoài) and automatic sleep stage classification could be a
healthy controls (HCs). Resting-state functional helpful assistance for experts on the analysis of sleep
magnetic resonance imaging (rs-fMRI) is a non- signals, the main motivation for the current work is
invasive technique, which could detect the the lack of a systematic method for automatic
ongoing neuronal process at the “resting state” artefact detection and cancellation which leads to an
through measuring the spontaneous brain activity improvement in the automatic stage classification
by low-frequency fluctuations in blood oxygen accuracy compared to the original acquired data
level-dependent (BOLD) signals, and
consequently provide a new opportunity to 5. MULTIPLE CLASSIFIERS FOR
investigate the functional abnormalities on SLEEPING STAGE CLASSIFICATION
several neurological disorders Classification is the process of categorizing
3. MOTIVATION data into relevant groups. The first step in the
classification process is the identification of features
The function of human body is frequently or characteristics that will enable the discrimination
associated with signals of electrical, chemical, or between the different groups of data. A classification
acoustic origin. Such signals convey information model should be developed in a way that provides a
that may not be immediately perceived because it structure for how the classification processes’
is hidden in the signal's structure. However, actions will be realized. Ideally, this model should
signals’ complexity is often considerable and be chosen to optimize the system performance,
therefore, the biomedical signal processing has although it may need to be revised as the classifier
become a vital tool for extracting clinically design progresses. A classifier is then implemented
significant information hidden in signals. The and “trained” to recognize the chosen features in the
artefacts such as body movements, sweating and data, or to determine the best input-to-output
sensor fault can reduce the accuracy in signal mapping. Generally, there are two ways to train a
processing especially in sleep signal analysis. The classifier: supervised learning and unsupervised
conventional solution is to detect the artefacts and learning. A system is called supervised learning if it
denoise the signal by removing corresponding uses data labelled by the expert to create an optimal
epochs from the sleep signal. However, this way, response for the system, which is used as feedback
the EEG signal will be manipulated and may lose to the learning system to increase accuracy. In
important information. One of the motivations of contrast, unsupervised learning occurs when the
this thesis is to develop and improve noise system does not use any labelled data to modify its
cancelation method that does not manipulate the output. Once the system has trained and learned, it
signal and protect its originality. Deep learning is is ready to recognize and classify specific inputs. It
an emerging technique that can be applied to a can be tested and evaluated with such metrics as
broad field of science in order to improve learning speed of computation and accuracy of classification.
and classification algorithms. Deep learning is
rarely used to classify bio signals and still there is
a lack of applying this technique to sleep staging

74
 International Journal of Scientific Research in Engineering and Management (IJSREM)
Volume: 07 Issue: 03 | March - 2023 Impact Factor: 7.185 ISSN: 2582-3930

5.1 CONVENTIONAL CLASSIFIERS components. This two-way classification scheme,

however, misses a key insight gained in deep
Several papers provide evidence for the learning research about how generative or
high performance of Support Vector Machine unsupervised-learning models can greatly
(SVM) specifically for high dimensional improve the training of DNNs and other deep
classification problems. In principle, SVMs are discriminative or supervised-learning models via
designed for binary classification problems better regularization or optimization. deep
(discrimination between two classes). However, learning has been proposed for unsupervised
as in many classification tasks, automatic sleep feature learning and has been applied to many
scoring requires discrimination between multiple domains, such as biomedical signals. One of the
classes (Awake, S1, S2, S3 and REM). Hence, for major advantages of deep learning compared to
getting benefit from the assumed advantages of traditional approaches is that they can work
SVM classification, a multi class SVM directly on raw data and do not require any tuning
framework needs to be implemented. Two of the or hand-crafted features. Instead, they can learn
most widely used approaches for multi-class their own feature representations.
SVM classification are the One-Against-All
(OAA) and the One-Against-One (OAO) 5.3 CONVLUTIONAL NEURAL NETWORK
approaches. The OAA framework consists of
using a binary SVM to distinguish each class from Convolutional Neural Network (CNN) A
all other classes and the decisions obtained by CNN is a multilayer perceptron designed
applying a winner takes-all strategy. SVM specifically to recognize two dimensional shapes
techniques are often proposed for anomaly with a high degree of invariance to translation,
detection and decision-making tasks in healthcare scaling, skewing, and other forms of distortion.
services. However, SVM is not an appropriate Learning section of this classifier is done in
method to integrate domain knowledge to use supervised method which includes the following
metadata or symbolic knowledge seamlessly with structure:
the measurements from the sensors. Moreover, • Feature extraction;
like other classifiers, SVM cannot be applied to • Feature mapping;
find the unexpected information from unlabelled • Subsampling;
data.
The weights in all layers of a CNN are learned
5.2 DEEP LEARNING APPROACHES through training. Also, the network learns to extract
its own features automatically
Unlike some of the machine learning
areas such as natural language processing and
object classification, the potential of deep
learning techniques is not fully explored in
automatic sleep stage classification. This fact is
also noticeable when it comes to the feature
transformation for sleep scoring. By the
commonly adopted machine learning tradition
naturally deep learning techniques classify into
deep discriminative/supervised models 81 (e.g.,
deep neural networks (DNNs), RNNs,
convolutional neural networks (CNNs), etc.) and
generative/unsupervised models (e.g., restricted
Boltzmann machine (RBMs), deep belief
networks (DBNs), deep Boltzmann machines Fig 2: Convolutional neural network
(DBMs), regularized autoencoders, etc.). The framework
third category belongs to the class of hybrid deep
network structures, which refers to the deep This diagram shows the multiple layers such as
architecture that either comprises or makes use of input layer, four hidden layers, and an output layer.
both generative and discriminative model This network is designed to perform image
processing. The input layer consists of 28×28

75
 International Journal of Scientific Research in Engineering and Management (IJSREM)
Volume: 07 Issue: 03 | March - 2023 Impact Factor: 7.185 ISSN: 2582-3930

sensor neurones, receives the images of different CONVOLUTIONAL NEURAL 70

characters that have been approximately centred NETWORK
and normalized in size. We can use the CNN
algorithm framework to classify multiple sleeping
stages. ACCURACY (%)
72

70

68

66

64

62

60

58

56

54

Fig 3: Proposed Work

The proposed work shown in fig 3 to analyse two
types of datasets such as CSV file and EDF
datasets. Then perform CNN algorithm to classify
the sleeping stage with improved accuracy rate.

Fig 3: Accuracy
6.EXPERIMENTAL RESULTS
From the diagram, convolutional neural network
In this study we can input the EEG provides the improved accuracy than the existing
datasets related to sleeping stages and developed frameworks
in Python framework. Accuracy as main criterion
is considered for evaluating and comparing the 7. CONCLUSION
different classification methods. The performance In this work, we presented an automatic
of the sleep stage classification is evaluated using sleep stage classification model that could achieve
repeated random sub-sampling validation. To good performance on the public dataset and
measure the classification accuracy, the overall accurately predict the sleep stage on our own
accuracy value is calculated as follows laboratory dataset. There are many solutions for
Number of true detections classification of sleep signals. In this paper, we
proposed two methods and explored their utility and
ACCURACY =
benefits in the study of sleep. Convolutional neural
Total number of epochs
network provides improved performance than the
existing machine learning and deep learning.
ALGORITHMS ACCURACY
(%)
SUPPORT VECTOR MACHINE 60
DEEP NEURAL NETWORK 65

76
 International Journal of Scientific Research in Engineering and Management (IJSREM)
Volume: 07 Issue: 03 | March - 2023 Impact Factor: 7.185 ISSN: 2582-3930

REFERENCES [10] Wen, Ze-ying, et al. "Identification of

discriminative neuroimaging markers for patients
[1] Malik, Asra, et al. "Detection of
on hemodialysis with insomnia: a fractional
Insomnia using Electrocardiography and
amplitude of low frequency fluctuation-based
Electromyography." 2021 International
machine learning analysis." BMC Psychiatry
Conference on Artificial Intelligence (ICAI).
23.1 (2023): 1-13.
IEEE, 2021.
[2] Yang, Buffing, and Hongxing Liu.
"Automatic identification of insomnia based on
singlechannel EEG labelled with sleep stage
annotations." IEEE Access 8 (2020):
104281104291.
[3] Kuo, Chih-En, and Guan-Ting Chen. "A
shorttime insomnia detection system based on
sleep EOG with RCMSE analysis." IEEE Access
8 (2020): 69763-69773.
[4] Andresini, Giuseppina, et al. "Insomnia:
Towards concept-drift robustness in network
intrusion detection." Proceedings of the 14th
ACM Workshop on Artificial Intelligence and
Security. 2021.
[5] Sharma, Manish, Harsh S. Dhiman, and
U. Rajendra Acharya. "Automatic identification
of insomnia using optimal antisymmetric
biorthogonal wavelet filter bank with ECG
signals." Computers in Biology and Medicine
131 (2021): 104246.
[6] Lee, Mi Hyun, et al. "Multitask fMRI and
machine learning approach improve prediction of
differential brain activity pattern in patients with
insomnia disorder." Scientific reports 11.1
(2021): 1-13.
[7] Afshani, Mortaza, et al. "Discriminating
paradoxical and psychophysiological insomnia
based on structural and functional brain images:
a preliminary machine learning study." (2022).
[8] Islam, Md Muhaiminul, et al. "Prediction
of chronic insomnia using machine learning
techniques." 2020 11th International Conference
on Computing, Communication and Networking
Technologies (ICCCNT). IEEE, 2020.
[9] Ma, Xiaofen, et al. "Functional
connectome fingerprint of sleep quality in
insomnia patients: Individualized out-of-sample
prediction using machine learning." NeuroImage:
Clinical 28 (2020): 102439.

77