sensors

Review
Electroencephalography Signal Processing: A Comprehensive
Review and Analysis of Methods and Techniques
Ahmad Chaddad 1,2, * , Yihang Wu 1 , Reem Kateb 3 and Ahmed Bouridane 4

1 School of Artificial Intelligence, Guilin University of Electronic Technology, Guilin 541004, China
2 The Laboratory for Imagery, Vision and Artificial Intelligence, Ecole de Technologie Supérieure,
Montreal, QC H3C 1K3, Canada
3 College of Computer Science and Engineering, Taibah University, Madinah 41477, Saudi Arabia
4 Centre for Data Analytics and Cybersecurity, University of Sharjah, Sharjah 27272, United Arab Emirates
* Correspondence: ahmadchaddad@guet.edu.cn or ahmad8chaddad@gmail.com

Abstract: The electroencephalography (EEG) signal is a noninvasive and complex signal that has
numerous applications in biomedical fields, including sleep and the brain–computer interface. Given
its complexity, researchers have proposed several advanced preprocessing and feature extraction
methods to analyze EEG signals. In this study, we analyze a comprehensive review of numerous
articles related to EEG signal processing. We searched the major scientific and engineering databases
and summarized the results of our findings. Our survey encompassed the entire process of EEG
signal processing, from acquisition and pretreatment (denoising) to feature extraction, classification,
and application. We present a detailed discussion and comparison of various methods and techniques
used for EEG signal processing. Additionally, we identify the current limitations of these techniques
and analyze their future development trends. We conclude by offering some suggestions for future
research in the field of EEG signal processing.

Keywords: EEG; signal processing; machine learning

Citation: Chaddad, A.; Wu, Y.; Kateb,

R.; Bouridane, A.
1. Introduction
Electroencephalography Signal Brain science has become an essential field of study to unravel the mysteries of life due
Processing: A Comprehensive to developments in biomedical technology and our growing understanding of the brain [1].
Review and Analysis of Methods and Since the middle of the 20th century, the complexity of the brain has been studied and, in
Techniques. Sensors 2023, 23, 6434. recent years, brain science has continued to be a hot topic for research [2]. The analysis of
https://doi.org/10.3390/s23146434 brain electrical activity is one of the main areas of interest in brain science [3]. As a result,
Academic Editors: Chang-Hwan Im the electroencephalogram (EEG) is important to analyze brain science and is often used in
and Yvonne Tran various brain-related research domains [4,5].
EEG is a noninvasive neuroimaging technique that involves the placement of elec-
Received: 21 May 2023 trodes on the scalp to record electrical activity of the brain [6]. This enables researchers
Revised: 16 June 2023
to measure and analyze the electrical signals generated by the brain. These signals offer
Accepted: 13 July 2023
valuable information on the operating mechanisms of the brain, covering the identifica-
Published: 16 July 2023
tion of various neurological disorders and the exploration of cognitive processes such as
perception, attention, and memory. EEG has gained widespread popularity as a means of
investigating electrical activity of the human brain, due to its noninvasive and safe charac-
Copyright: © 2023 by the authors.
teristics [7]. In addition, EEG signals have the potential to be integrated with other imaging
Licensee MDPI, Basel, Switzerland. modalities, including magnetic resonance imaging (MRI) [8,9], functional near-infrared
This article is an open access article spectroscopy (fNIRS) [10–12], and positron emission tomography (PET) [13], in order to
distributed under the terms and achieve a better understanding of brain function and structure.
conditions of the Creative Commons EEG is a signal pattern that is obtained by amplifying and recording the spontaneous
Attribution (CC BY) license (https:// biological potential of the brain on the scalp. This potential has been shown to reflect the
creativecommons.org/licenses/by/ macroscopic activity of the brain surface and is typically acquired using noninvasive elec-
4.0/). trodes applied onto the scalp. These electrodes capture the inherent and periodic electrical

Sensors 2023, 23, 6434. https://doi.org/10.3390/s23146434 https://www.mdpi.com/journal/sensors

Sensors 2023, 23, 6434 2 of 27

impulses generated by clusters of brain cells [14]. Currently, EEG is widely used in the
field of neuroscience and has the potential to advance brain–computer interfaces, facilitate
emotion detection, and help in partial paralysis rehabilitation [15,16]. Furthermore, EEG is
a valuable tool for clinicians and researchers in identifying brain dysfunction-associated
diseases, including but not limited to Alzheimer’s disease [17,18], epilepsy, schizophrenia,
Creutzfeldt–Jakob disease [19], cerebral palsy [20], and cognitive impairment [21].
Accurately identifying and analyzing EEG signals requires a thorough understanding
of their complex and theoretical properties, as well as the extraction of relevant features
for the given task. However, EEG signals pose significant challenges due to their unique
characteristics. According to [22], one such challenge is their susceptibility to noise interfer-
ence, resulting in a low signal-to-noise ratio. Additionally, their nonlinearity and lack of
conformity to a normal distribution distinguish them from conventional signals. Moreover,
individual factors such as age, psychology, and testing environment can cause significant
variations in EEG signals [23]. In [22], it is observed that the unique properties of EEG
signals pose a challenge in extracting pertinent information about specific tasks directly
from them. Therefore, it is imperative to develop various methodologies for signal analysis
and investigate machine learning techniques for signal analysis to better understand EEG
signals [24]. Accurate extraction of relevant information on specific tasks from EEG signals
requires careful consideration of their distinctive characteristics and the advancement of
sophisticated signal analysis methodologies. As emphasized in [25], precise detection and
analysis of EEG signals are crucial to advance our understanding of brain functioning.
The high interest in EEG as a research domain is apparent in the Google Scholar, PubMed,
and Web of Science search results obtained between 2016 and 2022, as illustrated in Figure 1.
Our paper presents a novel contribution through a comprehensive description of denoising
techniques, which includes mathematical formulations with pseudocodes. In addition, we
report the recent advancements in the field of EEG, while highlighting current challenges
and discussing future trends.
The primary contributions of this paper can be summarized as follows.
• We present a detailed examination of the EEG signal analysis process, including the
stages of signal acquisition, denoising, and feature engineering.
• The procedure used to denoise the EEG signal is described in full, along with the
accompanying evaluation standards.
• We examine feature engineering in detail in this paper, looking at time–frequency,
high-order spectral, and nonlinear dynamic analysis.
• We give a thorough analysis of both traditional and deep learning methods for catego-
rizing EEG signals. We also provide an overview of the typical datasets utilized for
EEG signal processing.
• We highlight current issues with EEG signal processing techniques and offer potential
solutions as well as future research prospects.
The structure of this paper is as follows. Section 2 presents a brief description of the
impact of EEG acquisition as a noninvasive biomedical device. It gives a full evaluation
of various denoising techniques, and an explanation of the merits and downsides of each.
Additionally, it examines the significance of feature engineering and its various techniques,
as well as the use of machine learning- and deep learning-based classifiers. In Section 3,
future perspective and limitations are discussed. The article ends in Section 4, where the
research efforts and contributions of this investigation are outlined.
Sensors 2023, 23, 6434 3 of 27

Figure 1. Bar graph illustrating the number of published papers over the years obtained from Google
Scholar, PubMed, and Web of Science searches for the subject of the electroencephalogram (EEG).
The search query used was “EEG” in the title.

2. The Pipeline of EEG Signal Analysis

In this section, the focus is on the methodology of EEG signal analysis, as illustrated
in Figure 2. The pipeline is related to the classification of EEG signals. The discussion
begins with an explanation of the methodology of EEG signal acquisition using equipment.
Then, the algorithm for denoising EEG signals is examined, which is capable of eliminating
erroneous data and extracting relevant information. Next, the feature engineering process
is discussed, which involves removing the less important features. Finally, the use of deep
learning and machine learning algorithms for classification is explored.

Figure 2. The four-step process for EEG signal analysis. The EEG signal analysis involves four stages:
acquisition, denoising, feature engineering, and classification.

2.1. Acquisition
EEG is a neurophysiological technique used to measure and quantify neural activity
in various regions of the brain. The brain consists of a large number of neurons and their
activities generate distinct scalp potentials, producing signals in different states of alertness,
response to external stimuli, and other factors unique to each individual [26]. To obtain
data sources for various applications and research on EEG signals, appropriate acquisition
equipment is essential. Acquisition of EEG signals can be classified into two primary
categories: invasive and noninvasive [27]. Invasive acquisition involves surgical insertion
of electrodes into the cerebral cortex or other regions of the brain to capture signals. On the
other hand, noninvasive techniques use EEG sensors positioned on the scalp’s surface
and do not require implanted electrodes. Currently, most EEG signal acquisition methods
are noninvasive. In Algorithm 1, we provide the main steps of EEG signal analysis.
During the initial stages of the acquisition of EEG signals, the German scientist Hans
Berger detected electrical signals in the cerebral cortex via a galvanometer in 1924. After pre-
liminary investigations, the scientists began implanting metallic electrodes directly into the
cerebral cortex to capture comprehensive EEG readings. With advancements in computer
technology, EEG signal collection techniques have improved significantly, leading to a
higher resolution of collected EEG signals. Most EEG signal acquisition instruments that
have reached relative maturity currently employ PC displays, wired data transmission,
Sensors 2023, 23, 6434 4 of 27

and external power sources. These devices exhibit robust data processing capabilities,
favorable outcomes, and consistent performance. However, they have a heavy form factor,
pose potential hazards, and exhibit high power consumption. Consequently, the need for
portable EEG collection equipment has been observed [28].

Algorithm 1 Pipeline of EEG signal analysis

Input: EEG signal eeg(t).

procedure Acquisition (eeg(t)):

Sampling equipment selection.
Choose invasive or noninvasive acquisition.
Choose the best reference electrode.
Return: EEG data eeg(t).

procedure Denoising (eeg(t)):

Using different denoising methods to process EEG signals.
Return: Denoised EEG signal eeg(t)new .

procedure Feature_Engineering (eeg(t)):

Perform time–frequency, high-order spectrum analysis or nonlinear analysis.
Return: More expressive features, eeg(t).

procedure Classification (eeg(t), choice):

Perform classification tasks based on choice:
If choice equal traditional approach do
Using KNN, SVM, ... to classify.
If choose the same deep learning approach do
Using CNN, GAN, ... to classify.
Return: Accuracy, precision, etc.

Portable EEG acquisition devices such as the Emotiv EPOC have gained popularity in
recent years. The Emotiv EPOC utilizes nonimplantable electrodes and comprises 14 electri-
cal data acquisition channels and two reference electrodes. Stytsenko et al. [29] found that
the Emotiv EPOC can collect real EEG data. Emotiv EPOC neural headphones are also avail-
able on the market to measure brain activity [30]. However, the performance of the Emotiv
EPOC headset devices, while capable of recording EEG data, is inferior to that of larger
devices [31]. In another study, Martins proposed a wearable EEG acquisition device and a
sleep inertia detection system of the data analysis platform [32]. The system is a monolithic
low-power with a low-noise analog front-end EEG acquisition system [33]. The system has
demonstrated high precision and high reliability, and flexible adjustment. Moreover, a new
waterproof, lightweight, and portable EEG acquisition device was proposed to acquire and
analyze the EEG signals of dolphins [34]. The device was designed to enable relatively
unrestricted EEG acquisition. Their acquisition device is equipped with customized suction
cups with embedded electrodes. It also incorporates a Bluetooth module to communicate
with the ground station. Furthermore, they used the portable Muse brain wave sensor
device for stroke identification [35]. The device follows the international 10–20 system
and utilizes four recording electrodes (AF7, AF8, TP9, TP10) and one reference electrode
(Fpz). In [36], they designed a high precision portable EEG acquisition system using the
CompactPCI platform to address the limitations of existing EEG acquisition systems, such
as high costs and limited accuracy. Aside from the previous methods, there are many ways
to collect EEG signals. For example, deep brain stimulation was performed through the
use of neural electrodes that were placed in specific target regions of the brain [37]. These
electrodes generate current or voltage through an implantable pulse generator. Further-
more, the MR signal has the ability to reflect both oxygen saturation and blood flow in the
brain [38]. It can reflect the activity of neurons and serve the purpose of functional imaging.
Sensors 2023, 23, 6434 5 of 27

Figure 3 shows a short segment of an EEG signal [39] recorded using a 14-channel
Emotiv Epoc device. These 14 EEG signals are specific locations on the scalp where
electrodes are placed to measure brainwave activity. In general, the frequency of an EEG
signal is measured in Hz and refers to the number of cycles that occur per unit time [40].
Specifically, there are five waves that correspond to five ranges of frequencies in EEG
signals [41], namely, Delta [0.5 to 4 Hz], Theta [4–8 Hz], Alpha [8–13 Hz], Beta [13–30 Hz],
and Gamma [>30 Hz]. Delta waves are related to slow-wave activity in the brain and deep
sleep; Theta waves are commonly observed during periods of relaxation and meditation;
Alpha waves are most noticeable when the eyes are closed; Beta waves are associated with
cognitive processing that is actively engaged and can be detected during tasks that require
significant attention; and Gamma waves are linked to advanced cognitive processing and
the merging of sensory information. In addition, each EEG channel is named according to
its location relative to the midline of the head and its distance from the forehead or back of
the head. Together, these channels allow researchers and clinicians to measure electrical
activity in various regions of the brain, and gain insight into cognitive processes such as
attention, memory, and emotion.

Figure 3. An example of 14 EEG signal channels, where the x-axis denotes time and the y-axis
represents the magnitude of the 14 signals [39]. These signals can be characterized by their frequency,
which refers to the number of cycles per second (Hz) of the electrical activity. These channels are
named AF3, F7, F3, FC5, T7, P7, O1, O2, P8, T8, FC6, F4, F8, and AF4, which correspond to the specific
electrode placements on the Emotiv Epoc equipment.

Table 1 presents a summary of recent studies on EEG signal analysis in different

application domains. In [42], commercial EEG equipment was used to collect signals to
analyze driver fatigue. A third-order Butterworth bandpass filter was used to remove
irrelevant information from the raw signals [42]. Another study by Gamage et al. utilized
the Emotiv EPOC to detect the driver’s emotions while driving and extracted features
using EEGlab and other tools in Matlab. The data was then classified to develop an emotion
classification model aimed at reducing the risk of car accidents. The study demonstrated
the high reliability of the dataset collected by the device [43]. Table 1 provides relevant
details, such as the study objective, data source, and data processing techniques used in the
different studies.
Sensors 2023, 23, 6434 6 of 27

Table 1. Research on EEG applications.

Ref. Prupose Acquisition Method Data Processing

Commercial Neuroscan system with Third-order Butterworth bandpass
Ma [42] Recognize driver fatigue
40 electrodes filter
Detect driver’s EEG to reduce traffic Evoke the emotions of the test driver
Gamage [43] EEGLAB Toolbox of Matlab
accidents with video and audio
Traditional 128-electrode mounted
Strengthen the depression
Shen [44] elastic cap and a wearable 3-electrode EEGLAB Toolbox of Matlab
recognition performance
EEG collector
Detect the working status of
Saedi [45] Investigate mental and motor work A mix of macro and micro scrutiny
construction workers
Estimating classification accuracy
Han [46] Classification of eye state EEG measured around the ear
using 3 CNN models
Classic algorithms and the neural
Pawuś and Paszkiel [47] Use BCI to control the robot Emotiv EPOC
network
Chen [48] EEG decoding Obtained in the open world Supervised deep learning
Pei [49] PreG electrode in BCI Obtained form PreG electrode SSVEP-based BCI
Jemal [50] Epileptic seizure prediction Publicly available CHB-MIT dataset Deep neural network model
From 7 subjects participating in the
Wen [51] Evaluate spatial cognitive ability Coupling strength calculation
game
Li [52] Emotion recognition SJTU Emotion EEG Dataset Experiment-level BN
Freismuth et al. [53] Treatment and diagnosis of ADHD Wearable EEG device HiLCPS framework
CNN: Convolutional neural network; BCI: Brain–machine interface; BN: Batch normalization; ADHD: Attention
deficit hyperactivity disorder; HiLCPS: Human-in-the-loop cyber-physical systems.

2.2. Denoising
As mentioned above regarding the acquisition of EEG signals, multiple electrodes
are placed on the scalp. However, external interference can cause diverse artifacts to
emerge, which can compromise the quality of the signals. Physiological artifacts, such as
involuntary eye movements, blinking, heart activity, and muscle movement, are known to
be present in EEG signals and can negatively affect their quality [54]. Therefore, denoising
EEG signals has become a topic of significant research interest and attention. To ensure the
reliability of features extracted from EEG signals, it is essential to remove any associated
artifacts. Currently, several denoising techniques have been developed.

2.2.1. Regression Method

The traditional approach to remove eye artifacts in EEG signals is the regression-
based analysis approach. During the recording of the EEG signal, an electro-oculogram
is recorded concurrently to obtain coefficients for various noise sources such as blinking
artifacts (VEOG), eye movement artifacts (HEOG), and other noise sources, by using
regression analysis. These coefficients estimate the ratio of artifacts in a particular EEG
channel [55]. In [56], they proposed a general lagged regression model to represent this
process as follows.
eeg(t) = EEG (t) − ∑ Tg=0 β g eog(t − g) (1)
where eeg(t) and eog(t–g) are the recorded EEG and EOG information at times t and t˘g,
respectively. EEG (t) denotes the uncorrupted EEG at time t. The regression coefficient β g
measures the effect of EOG on eeg(t) at time (t˘g).
For example, to remove the effect of an artifact signal Y from an EEG signal X, we
compute the regression correlation between the two signals. This denoising function allows
us to successfully eliminate the artifact signal from the EEG. Algorithm 2 illustrates the
general procedure.
However, the main concern raised about the regression method is bidirectional con-
tamination [55]. For example, EOG recordings may include neural potential along with
ocular potential [57,58]. Therefore, removing EOG activity from the EEG signal requires
subtracting a portion of the relevant EEG signal from each recording. Furthermore, one
of the challenges of regression techniques is that they may not be effective in dealing
with other artifacts, such as EMG artifacts [59], due to the lack of clear reference channels.
With the emergence of potentially more efficient algorithms such as principal component
Sensors 2023, 23, 6434 7 of 27

analysis (PCA) and independent component analysis (ICA) [60,61], the regression method
is no longer the default choice for removing artifacts from an EEG caused by EOG or
ECG signals.

Algorithm 2 Regression-based denoising of EEG signals

Input: EEG signal X, artifact signal Y
Output: Clean EEG signal Z
function R EGRESSION(X, Y)
Calculate regression coefficients between X and Y
Remove artifact from EEG signal
return Clean EEG signal Z
end function

2.2.2. Blind Source Separation

Blind Source Separation (BSS) is a technique that separates source signals from a mixed
signal without prior knowledge of the original signals. In the first step, the observed EEG is
decomposed into its sources using the BSS algorithm. The noise sources are then identified
and eliminated while preserving the sources of brain activity. BSS techniques are commonly
used to denoise EEG signals [62]. The following techniques are described below.
Independent Component Analysis: Various types of ICA algorithms have been pro-
posed in different papers, making it one of the key blind source separation techniques in
biomedical engineering. ICA is capable of extracting statistically independent sources from
a collection of recorded signals [63]. In general, denoising EEG signals using ICA can be
expressed as follows [64]:
X = AS (2)
where the given equation involves three matrices: X, A, and S. Matrix X contains EEG data,
matrix A represents the linear mixing of various sources (e.g., EEG and artifact sources),
and matrix S consists of independent components, such as brain and artifact sources.
After obtaining the independent components, they can be visually examined to identify
any artifacts such as eye blinks and muscle activity. The EEG signals can be denoised by
removing these components. Algorithm 3 presents the denoising EEG signals based on the
ICA method.

Algorithm 3 ICA based denoising of EEG signals

Input: X: EEG data matrix
Input: n_components: number of independent components to estimate
Output: S: matrix of independent components
Output: A: estimated demixing matrix
Center and whiten the X.
Initialize A randomly.
repeat
Update A by exploiting non-Gaussianity of independent sources.
until convergence
Compute S from A and X.
Identify artifact components in S.
Remove artifact components from S.
Reconstruct cleaned data from S.
return S, A

Due to the significant overlap between EEG signals and EMG artifacts in both spatial
and temporal domains, conventional ICA algorithms often struggle to separate all EMG
artifacts and generate a set of independent components. To address this problem, Li et al.
introduced an improved ICA model called EMG removal by adding sources of EMG
(ERASE) [65]. Specifically, EMG reference artifacts were involved from the head and neck
Sensors 2023, 23, 6434 8 of 27

muscles as input to the ICA, which increased the power of the EMG artifacts in a few
independent components, resulting in a more accurate separation. In comparative tests,
the ERASE algorithm removed an average of 26% more EMG artifacts from EEG signals
than the traditional ICA algorithm. In [66], a new denoising framework was devised
and merges ICA with the continuous wavelet transform (CWT). CWT and the K-means
algorithm are used to map the detected blink artifact. EEG information is then preserved
while denoising through ICA [66].
Principal Component Analysis: PCA, which stands for principal component analysis,
is an easy-to-use data reduction technique that uses the principle of orthogonality to
eliminate artifacts [54]. Through the utilization of PCA dimensionality reduction, it is
possible to eliminate the presence of noise, represented by small eigenvalues, within the
data. This process results in a partial denoising effect. Typically, given EEG data X, the main
goal of PCA is to solve this equation:

XX T ωi = λi ωi (3)

where the symbol λ denotes the eigenvalue, while ω represents the eigenvector. The tech-
nique for reducing the number of dimensions in a dataset is accomplished by decomposing
the eigenvalues of the matrix XX T . The resulting eigenvalues are then sorted and the top
d values are selected to serve as a projection matrix. Subsequently, the projection matrix
can be used to transform the EEG data denoted by D into new EEG data represented
by D ∗ = W ∗T D, while minimizing the presence of noise. The pseudocode for PCA is
presented in Algorithm 4.

Algorithm 4 Typical principal component analysis

Input: EEG data D = { x1 , x2 , ..., xn }, low-dimensional space dimension d.
Output: Projection matrix W ∗ = (w1 , w2 , ..., wd ).

procedure PCA(D):
Sample centering xi ← xi − m1 ∑im=1 xi .
Calculate XX T .
Eigenvalue decomposition for XX T .
Select the largest d eigenvalues.
W ∗ = (w1 , w2 , ..., wd ).
New EEG data D ∗ = W ∗T D.
Return: D ∗ .

In recent times, there has been an increasing focus on utilizing principal component
analysis in conjunction with other techniques to achieve EEG denoising. Patel et al. have
demonstrated the effectiveness of combining ensemble empirical mode decomposition
(EEMD) with PCA to efficiently detect and suppress artifacts in single-channel EEG data.
This method can automatically detect and suppress eye artifacts after correct selection of
the detection threshold [67]. In [68], a learning model based on PCA and semi-supervised
support vector machine (SVM) is proposed. The model first preprocesses the EEG and
uses PCA to reduce its dimensionality. After obtaining a set of optimal channel subsets,
a semi-supervised classification model based on SVM is designed. This model determines
the relationship between labeled data and unlabeled data by calculating the Euclidean
distance between them, and then extracts features to identify them. The experimental
results indicate that the method can achieve 84.3% correct classification results with only
40% labeled data, suggesting its potential in scenarios where only a small amount of labeled
data is available [68].

2.2.3. Canonical Correlation Analysis

In the context of EEG signals contaminated with muscle artifacts, canonical correlation
analysis (CCA) is generally more effective than ICA [69]. Due to the relatively lower
Sensors 2023, 23, 6434 9 of 27

autocorrelation of muscular artifacts compared to brain activity, it is feasible to employ

canonical correlation analysis (CCA) as a means of distinguishing between muscle activity
and brain activity [70]. The common CCA formula used for denoising EEG signal is
expressed as follows [71]:
u T R xy v
max p q (4)
u
u T R xx u v T Ryy v

where the variables u and v represent canonical variates, which are linear combinations
of channels. The EEG data and artifact data have covariance matrices denoted by R xx
and Ryy , respectively. The cross-covariance matrix between the EEG data and artifact
data is represented by R xy . After obtaining the canonical variates, they can be used to
eliminate artifacts from the EEG data by subtraction. The denoising steps based on CCA
are presented in Algorithm 5.

Algorithm 5 CCA based denoising of EEG signal [72]

Input: X: EEG data matrix
Input: Y: matrix of auxiliary variables (e.g., EOG or ECG data)
Output: Z: matrix of cleaned EEG data
Center and whiten the X.
Initialize the weight matrices A and B randomly.
repeat
Compute the canonical weights wa by maximizing the correlation between X and Y
with respect to A. . wa : weights used to linearly combine the EEG signal for one
component
Compute the canonical weights wb by maximizing the correlation between X and Y
with respect to B. . wb :weights used to linearly combine the auxiliary signal for one
component
Update the weight matrices A and B.
until convergence
Compute the cleaned data as Z = A T X.
return Z

In [69], P Sheoran et al. proposed a new algorithm that combined CCA and noise
adjusted principal component transform (NAPCT) to eliminate noise in EEG data. Using
CCA to estimate the noise covariance matrix and NAPCT to remove artifact components,
the algorithm achieved this without human intervention [69]. Another study introduced
an unsupervised automated eye artifact recognition and removal algorithm [73]. This
algorithm used CCA to extract neural signals from data and used a multi-channel Wiener
filter (MWF) to adaptively eliminate eye artifacts from multi-channel EEG data [73].

2.2.4. Wavelet Transform

The signals recorded by EEG devices often contain irregularities. To analyze these non-
stationary signals, the wavelet transform (WT) is a widely used method [74,75]. The con-
ventional approach of WT divides the EEG signal into wavelet components. Components
that contain artifacts are identified and removed, leaving only clean components. These
clean components are then used to reconstruct a purified signal [76].
The WT is generally classified into two categories: discrete wavelet transform (DWT)
and continuous wavelet transform (CWT) [77]. Given the continuous nature of the EEG
signal, our attention is directed towards the DWT transformation. The DWT can be
expressed in mathematical notation:
R +∞
DWT (m, n) = −∞ x (t)ψm,n (t)dt (5)

where the variables “m” and “n” represent the scaling and translation factors, respectively.
Sensors 2023, 23, 6434 10 of 27

The process of discrete wavelet analysis involves the decomposition of x (t) into
various scales:
K ∞ ∞
x (t) = ∑ ∑ d j (k )ψj,k (t) + ∑ aK (k)φK,k (t) (6)
j=1 k =−∞ k =−∞

where the given equation involves discrete analysis wavelets represented by ψj,k (t) and
discrete scaling functions represented by φK,k (t). The variable d j (k ) denotes the detailed
signals or wavelet coefficients at a scale of 2 j , while aK (k) represents the approximated
signal or scaling coefficients at a scale of 2K .
DWT is a method of transforming time domain EEG signals without redundancy,
which is useful in removing artifacts. To accomplish this, the signal undergoes a series
of low-pass and high-pass filters to obtain approximate and detailed coefficients. This
process is repeated until the desired frequency is achieved. In Algorithm 6, we present
the pseudocode for DWT. However, DWT has the drawback of lacking translation invari-
ance. The stationary wavelet transform (SWT) can overcome this issue, but it has its own
limitations, such as redundancy and slow speed [78].

Algorithm 6 DWT based denoising of EEG signal [79]

Input: X: EEG data matrix (rows represent the EEG channels)
Output: Y: matrix of cleaned data
Set the wavelet basis and level of decomposition
for each channel c in X do
Compute the DWT coefficients of c at each level using the fatigue wavelet basis.
Identify the approximation coefficients at the desired level as the artifact-free signal.
Threshold the detail coefficients using a soft or hard thresholding technique.
Reconstruct the cleaned signal by inverse DWT using the modified coefficients.
Store the cleaned signal in the corresponding row of Y.
end for
return Y

WT alone may not be sufficient to address all the issues associated with EEG signal
denoising, as it can result in information loss and signal reconstruction problems. Therefore,
the combination of the wavelet transform with other techniques to improve the denoising
process has been explored. For instance, the authors of [80] applied ICA to separate signals
based on WT and found that this combination was effective in removing EMG noise and
ECG artifacts from EEG signals. In addition, notch filters can be used in conjunction with
WT to address the issue of overlapping spectra/frequencies between EEG signals and
artifacts. In [81], an adaptive threshold for wavelet coefficients was used to eliminate
frequent ocular artifacts (OA) and a 50 Hz IIR notch filter to reduce artifacts and noise
while preserving the original brain signals.

2.2.5. Empirical Mode Decomposition

Empirical mode decomposition (EMD) is a technique for analyzing non-stationary
and nonlinear signals that offers several desirable properties. EMD leverages the signal
extreme points to decompose a signal into a set of intrinsic mode functions (IMFs) and a
monotonic residual, which can be expressed using the following formula [82]:
n
x (t) = ∑ Ci (t) + r N (t) (7)
i =1

where Ci represents an IMF and r N represents the monotonic residual.

The IMFs are capable of capturing the fundamental oscillatory components at various
frequencies, which facilitates the differentiation between artifacts and the intended EEG
signal. To obtain the high-frequency content of a signal, including any artifacts, as a detail
component, one can subtract the envelope from the input signal. The envelope represents
Sensors 2023, 23, 6434 11 of 27

the smooth curve that passes through the local maxima and minima of the signal. This
technique enables the elimination of undesired artifacts while retaining the original EEG
signal. The reconstructed signal is obtained by adding up the detailed components after
cleaning. More details are listed in Algorithm 7.

Algorithm 7 Empirical Mode Decomposition (EMD) for EEG Artifact Removal [82]
Input: X: EEG data matrix
Output: Y: matrix of cleaned data
Set the stopping criterion and number of maximum iterations.
for each channel c in X do
Initialize d0 = c, k = 1 . dk : signal at iteration k
repeat
Find the local maxima and minima of dk−1 .
Compute the envelope by interpolating the maxima and minima.
Subtract the envelope from dk−1 to obtain the detail component hk . . h: detail
component
Update dk = dk−1 − hk .
Increment k.
until stopping criterion or maximum iterations are reached.
Compute the reconstructed signal as rc = ∑ik=1 hi . . rc : reconstructed signal for
channel c
Store the cleaned signal in the corresponding row of Y.
end for
return Y

One of the advantages of EMD is its ability to extract local amplitude, phase, and fre-
quency content from the resulting components. EMD is also adaptive and efficient and,
when combined with other techniques, it can lead to new advancements in the denoising
of EEG signals. For instance, combining ensemble empirical mode decomposition (EEMD)
with the CCA technique led to feasible results. Specifically, EEMD generates a large num-
ber of IMFs, increasing the number of channels available for ICA. This method leverages
interchannel information and addresses the challenging problem of CCA in dealing with
EEG data with low signal-to-noise ratio (SNR) and complex contamination [83].
EMD is highly sensitive to spike noise because of its reliance on extreme signal point
features for IMF decomposition. This sensitivity can lead to the mode-splitting effect, which
can seriously affect the removal of EOG artifacts. To address this issue, the multivariate
adaptive moving average–empirical mode decomposition (MAMA-EMD) based method
extracts peaks into the first IMF to improve the accuracy of subsequent IMF screening and
alleviate the mode-splitting effect [84]. However, MAMA-EMD may not achieve optimal
results in separating spikes when the pulse has two or more consecutive spike points.
To address this limitation, a new version of MAMA-EMD is proposed by supplementing
the minimum arc length criterion. This approach effectively eliminates the influence of
multi-point spikes on the screening process [85].
In recent years, various classic and commonly used denoising methods have been
combined to achieve better signal denoising in different situations. Some examples of these
methods or their combinations that have been used for denoising purposes are summarized
in Table 2.

2.3. Evaluation Criteria for Denoising

Various metrics, such as mean squared error (MSE), root mean squared error (RMSE),
signal-to-noise ratio (SNR), and percentage root mean square difference (PRD), are com-
monly used to assess the effectiveness of EEG signal denoising [86]. MSE is frequently
employed to evaluate the similarity between the initial EEG signal and the noise-reduced
signal. The RMSE is mathematically defined as the square root of the MSE. The SNR is a
metric used to compare the magnitudes of the signal and noise power. The PRD is used to
Sensors 2023, 23, 6434 12 of 27

measure the degree of similarity between the original and noise-reduced signals, with a
lower PRD indicating a higher degree of similarity between the two signals. These metrics
can be defined as:
s
n n
1 2 1 2
MSE = n ∑ (yi − ŷi ) , RMSE = n ∑ (yi − ŷi )
i =1 i =1
 n
∑ y2i
 (8)
n |y −ŷ |
100%
SNR = 10log10  n
i =1
2
, PRD =
n ∑ i|yi | i
∑ (yi −ŷi ) i =1
i =1

where yi typically refers to the true EEG signal value at time point i, ŷi represents the
denoised value of the EEG signal at time point i, and n is the total number of values.

Table 2. Application of denoising method in EEG signal.

Ref. Signal Processing Method Conclusion

More precise EMG separation without manual
Li [65] EMG reference artifacts of neck and head muscles
intervention
It is suitable for situations with few EEG signal channels
Maddirala and Veluvolu [66] CWT and K-means
and can accurately separate artifacts
Automatic detection and suppression of human eye
Patel [67] Combining EEMD and PCA artifacts can be achieved

It is suitable for processing signals with a low

PCA with an SVM-based semi-supervised
Xie [68] signal-to-noise ratio and only a few labels, with high
classification model
recognition accuracy and less training time
Artifact components are removed without manual
Sheoran and Saini [69] Combining CCA and NAPCT
intervention
Eye artifacts can be adaptively removed from
Miao [73] CCA and MWF multi-channel EEG data without the need for a reference
signal
The combination of wavelet transform and ICA can
Zhou and Gotman [80] Wavelet transform
effectively remove EMG and ECG artifacts in EEG signals
Effectively reduces artifacts and noise while preserving
Tibdewal [81] Use the adaptive threshold of wavelet coefficients
the original brain signal
It can make good use of interchannel information and
Chen [83] EEMD and CCA techniques has a good artifact removal effect in the case of serious
signal pollution
Can alleviate splitting effects, but not suitable for
Yang [84] Extract spikes to the first IMF
separating multipoint spikes
It can eliminate the effect of multipoint spikes on IMF
Li and Zhang [85] EMD
screening and better remove EOG artifacts
ICA: Independent component analysis; PCA: Principal component analysis; CCA: Canonical correlation analysis;
WT: Wavelet transform; EMD: Empirical mode decomposition; CWT: Continuous wavelet transform; EEMD:
Ensemble empirical mode decomposition; NAPCT: Noise adjusted principal component transform; MWF: Multi-
channel Wiener filter.

2.4. Feature Engineering

Enhancing feature engineering can improve the accuracy of predictions made on raw
data by transforming them into more expressive features. Extracting the available features
from the processing of EEG signals is a complex task that typically requires multiple human
experts with specialized knowledge. Machine learning techniques, such as deep neural
networks and adversarial generative networks, have allowed the automated extraction of
features from EEG signals. However, the interpretability problem of deep learning is deeply
criticized. Recently, the advancement of explainable AI (XAI) methods [87] has aimed at
improving the interpretability of deep learning models. For example, XAI based Smooth-
Grad [88] is used to perform EEG based emotion recognition [89], seizure detection [90],
and other applications. Using XAI methods, it allows eliminating the need for manual
selection by human experts. In [91], the authors proposed a method to achieve high-
resolution assessment of neural activity using deep networks, involving the implementation
of relevant layer propagation. They used an adversarial generation network to produce
EEG signals [92]. However, efficient feature engineering methodologies can help machine
Sensors 2023, 23, 6434 13 of 27

learning models acquire the fundamental features implicit in EEG more easily. Therefore,
we examine the main feature engineering methodologies used in EEG [91–93]. Commonly
utilized conventional signal processing techniques in various research studies on EEG signal
processing include time–frequency analysis, high-order spectrum analysis, and nonlinear
dynamics analysis.

2.4.1. Time–Frequency Analysis

The primary purpose of time–frequency analysis is to establish a link between the time
and frequency domains. This involves analyzing and processing signals in both domains
to extract relevant features. The most commonly used methods for analyzing stationary
signals include analysis of variance, waveform parameter analysis, wave identification,
histogram analysis, correlation analysis, and others. These methods are often applied in
the diagnosis of diseases [94]. For example, time–frequency analysis is used to map EEG
signals in the time and frequency domains. By dividing the signal data into windows and
scoring them, it is possible to identify epilepsy signals in the time–frequency domain [94].
Another example is related to the detection of peak features in EEG signals [95]. Specifically,
the signal is denoised using SVD and then the peak is detected, which yields superior
results [95]. In [96], a high resolution non-parametric time–frequency method is proposed
to analyze EEG signals that uses CNN to optimize the Wigner–Ville distribution as an input
without parameters, showing its clear superiority over other methods.

2.4.2. High-Order Spectral Analysis

Although the time domain analysis method falls short in analyzing high-order informa-
tion and providing complete signal feature results, the high-order spectral analysis method
can effectively address this limitation [97,98]. With its ability to map specific information
more effectively in EEG signal processing, higher-order spectral analysis demonstrates
significant superiority [97]. It is capable of suppressing Gaussian noise and producing
spectral structures that reflect more information, as evidenced by simulation experiments
on more than 200 EEG samples conducted by some authors [97]. The resulting spectral
lines are flatter, with less noise and smoother contours [97]. To identify the nonlinearity
and high dimensionality present in epileptic signals, the principal component features
are extracted using PCA on the 15 high-order spectra (HOS) features extracted from the
EEG data [99]. Furthermore, in [100], they use HOS to analyze EEG signals in the field
of neuro-marketing. According to the study findings, the proposed model, using SVM
with Gaussian kernel, achieved an average accuracy of 73.24% across all users. Moreover,
HOS features were used to access participants who were in a typical emotional state but
not exhibiting any motor movements [101]. Experiments demonstrated that the method
achieved an average accuracy of 95.7%.

2.4.3. Nonlinear Dynamic Analysis

Previous work has shown that the traditional linear analysis method is inadequate
for accurately evaluating the dynamic structure of EEG signals and as a result it cannot
reveal the essential characteristics of brain activity. However, the use of nonlinear dynamic
methods to extract and analyze EEG signals has provided a new approach to further study
the process and characteristics of human brain activity [102].
Much literature in this field uses nonlinear techniques, such as the Lyapunov expo-
nent, complexity measures, and fractal dimension, to analyze EEG signals. For example,
a researcher used the correlation dimension of nonlinear dynamics and the Lyapunov
index to extract characteristics of high-frequency EEG from elderly and young subjects
during various activities, including silent eye closure, silent eye-opening, and N-back letter
memory events. They then performed a statistical analysis on the resulting eigenvalues to
compare the differences between the two groups of eigenvalues [103].
The features of the EEG signals were effectively extracted by various indicators such
as the Hurst index, the Lyapunov index, the sample entropy, and the wavelet entropy [104].
Sensors 2023, 23, 6434 14 of 27

Furthermore, an adaptive Lempel–Ziv complexity algorithm was proposed and utilized to

measure the complexity of EEG signals, which was capable of identifying emotions [105].
By comparing the complexity values of the traditional Lempel–Ziv–Welch compression
algorithm (LZC) and the multiscale and adaptive LZC algorithm on the corresponding
electrodes under three emotional states, it was discovered that the adaptive LZC algorithm
could effectively distinguish between the three different emotional states. The algorithmic
processes of permutation entropy and sample entropy were briefly introduced and their
respective advantages and disadvantages were analyzed in detail [106]. A new algorithm
called equal symbolic entropy (ESE) was proposed, and its effectiveness in terms of accuracy
and efficiency was verified through simulation. It was applied to analyze emotional human
EEG signals from an experimental group [107].
Furthermore, due to the intricate nature of the EEG signal, traditional linear techniques
face significant challenges when analyzing it, leading to the application of nonlinear
dynamics methods in the analysis of EEG signals. Various features such as the correlation
dimension, the fractal dimension, the complexity, the approximate entropy, and others
have been explored in the literature, along with the corresponding methods for extracting
these features from EEG signals. These findings provide evidence supporting the scientific
basis of utilizing nonlinear dynamics for EEG signal analysis [108].

2.5. EEG Based Classifications

EEG signal classification is a fundamental task in the analysis of brain function, which
can be considered as one-dimensional biomedical signal processing [109–111]. Various
processing methods can be employed to classify EEG signals, including statistics, machine
learning (deep learning), and other techniques. In this section, we will focus on the
classification methods used in various fields of EEG application, with particular emphasis
on machine learning-based approaches. Furthermore, we have listed the commonly used
EEG datasets in Table 3.

Table 3. Summary of commonly used EEG signal analysis datasets.

Dataset Sample (n) Types SF (Hz)

Zhang [112] 122 Object recognition 256
Koelstra [113] 32 Emotion analysis 128
Zheng and Lu [114] 15 Emotion recognition 200
Ang [115] 9 Emotion recognition 250
Tangermann [116] 9 BCI 250
Sajda [117] 9 BCI 100
Andrzejak [118] 10 Seizure detection 173.86
Shoeb [119] 23 Seizure detection/prediction 256
Detti [120] 14 Seizure detection/prediction 512
Zhang [121] 6 Mental workload 500
Venkatachalam [122] 5 MIC 150
Zhang [123] 64 EEG denoising 512
BCI: Brain computer interface; MIC: Motor imagery classification; SF: Sample frequency.

2.5.1. Traditional Classification Method

The classification of brain signals through the application of ML techniques mainly
involves the use of supervised and unsupervised learning methods. These methods include
naive Bayesian (NB), decision tree (DT), K-nearest neighbor (KNN), support vector machine
(SVM), and random forest (RF), among others. Supervised learning is related to using input
and anticipated output data to develop predictive models that cater to classification and re-
gression. On the other hand, unsupervised learning involves proposing a prediction model
that uses input data for clustering and dimension reduction [124]. Supervised learning
yields higher accuracy than unsupervised learning when using classifiers such as SVM or
KNN. The precision of a solitary classification technique is restricted to particular use cases.
Hence, multimodal integration algorithms are commonly employed in studies to improve
the overall accuracy of classification. The increasing complexity of algorithms may result
in bias and affect their accuracy. Previous research has used machine learning methods to
Sensors 2023, 23, 6434 15 of 27

examine EEG signals for the identification of diseases (e.g., epilepsy, depression, stroke)
and rehabilitation interventions (e.g., motion imagination). Specifically, Table 4 summa-
rizes literature examples related to epilepsy, motion imagination, depression, and stroke.
In general, the results in Table 4, suggest that the SVM model is an important classifier
model for EEG signals.

Table 4. Application of traditional classification method in EEG research.

Ref. Domain Propose Method Conclusion

Used KNN and ANN classifiers to predict For KNN classifier, HFD with sample entropy had the
H and A. [125] Epilepsy
seizures highest accuracy of about 98%
Created an SVM classifier based on nonlinear fea-
Ping [126] Epilepsy Successfully improved the correct recognition rate
ture extraction
RF had better results in eliminating epilepsy error detec-
Jamunadevi [127] Epilepsy Used RF for detection and evaluation
tion
Added Gaussian noise and performed binary The maximum average classification accuracy of KNN
Jiahui [128] MI
classification classifier reached 88.57%
Created a lower extremity MI classification algo- The average classification accuracy of the two paradigms
Jiaying [129] MI
rithm based on LDA+KNN. was 67.5% and 84.62%, respectively
Dongare and Padole Created a majority voting classifier that combines
MI The accuracy of performance measurement was 85.36%
[130] SVM, LDA, and ANN
Ren [131] Stroke Adopted C4.5 decision tree Constructed a DT model with 37 nodes
Huaiwen and Yin [132] Stroke Used ROC and AUC for model screening The SVM model performed best as AUC = 1.000
Build model based on LASSO, DWI, PWI, and The accuracy of the combined model was 0.822, which
Hanqi [133] Stroke
SVM was better than that of the single sequence model
The random forest model had the best prediction effect,
Used CTA image collection set data and K-fold
Yong [134] Stroke with an accuracy of 94.9% and 90.8% in predicting new
cross validation
ischemic stroke
DWT: Discrete wavelet transform; SVM: Support vector machines; KNN: K-nearest neighbor; ANN: Artificial
neural network; HFD: Higuchi fractal dimension; RF: Random forest; MI: Motor imagery; LDA: Linear discrim-
inant analysis; FIR: Finite impulse response; SSA: Sparrow search algorithm; K-S: Kolmogorov–Smirnov; NB:
Naive Bayesian; MF-DFA: Multifractal declined fluctuation analysis; DT: Decision tree; ROC: Receiver operating
characteristic; AUC: Area under ROC curve; DWI: Diffusion weighted imaging; PWI: Perfusion weighted imaging;
CTA: Computed tomographic arteriography.

We also summarize many publications in Table 5, related to EEG signals to evaluate

many other tasks. For example, ANN and KNN classifiers are used to classify the EEG
signals into epileptic and non-epileptic classes. The results showed that the combination
of statistical parameters and classifiers achieved an accuracy of 95.9% for ANN and 92.4%
for KNN, indicating the effectiveness of the proposed method in detecting epilepsy using
EEG signals [125]. EEG with an SVM classifier based on nonlinear feature extraction is
used to improve the recognition rate of epileptic brain signals. They decomposed the
EEG signal into different frequency bands through a four-layer WT and the approximate
entropy (ApEn) value of the wavelet coefficients in each frequency band was used as the
eigenvector input. The correct recognition rates for normal and epileptic EEG signals
were 98.3% and 95.6%, respectively, which outperformed other similar algorithms [126].
Furthermore, the results showed that RF performed better than AdaBoost and KNN in
eliminating error detection [127].
Furthermore, there is extensive research on classifying EEG signals in motor imagery
(MI), a process that involves imagining actions without physically executing them. This
research is important in helping patients who have lost motor function in recovery. In [128],
it proposed to add Gaussian noise to EEG signals to improve recognition rates and perform
binary classification (left hand and right hand) MI tasks. The KNN algorithm achieved a
maximum average classification accuracy of 88.57% [129]; the LDA algorithm was used
to reduce the dimension of the feature data in the classification of MI EEG signals. When
combined with KNN, they achieved average classification accuracies of 67.5% and 84.62%,
respectively. This improved classification accuracy and speed, demonstrating the algo-
rithm’s advantages in lower-limb MI classification. In [130], the mixed feature majority
vote classifier is used to recognize MI EEG signals. They selected the combined feature
as the input of the classifier and used most voting classifiers to combine SVM, LDA,
Sensors 2023, 23, 6434 16 of 27

and ANN. The accuracy of performance measurement reached 85.36%, indicating that
the proposed system outperformed the conventional machine learning EEG recognition
classifier. In [135], they expanded the use of BCI to include motor imagery and presented a
framework that used augmented covariance extracted from an autoregressive model for
classification purposes.

Table 5. Summary of various EEG studies in a variety of applications.

Ref. Domain Proposed Method Conclusion

Proposed Hamming window bandpass FIR
The RF classifier based on negative emotion
Alharbi and Alotaibi [136] GD filter model for automatic gender
EEG signal had the lowest error percentage
identification using classifiers
Evaluated the performance of the nonlinear The maximum accuracy rate of RBF kernel
Parmar [137] Dyslexia kernel of SVM Gaussian kernel (RBF), for nonverbal stimuli reached 62.4%, with
polynomial kernel, and sigmoid kernel good performance
Constructed a multi-class SVM classifier The highest classification accuracy reached
Ling and Aihua [138] BCI
combining DT and SVM 80.8%
The accuracy of KNN was 81.6% better than
Used a new method to decode English
Hossain [139] CR SVM and NB in the classification of EEG
letters directly from EEG signals
signals with different letters
SVM provided the best performance with a
Design of a suitable classifier to distinguish correct classification rate of 90.14% for SZ
Padayatty and K [140] Schizophrenia
SZ EEG signals from HC EEG signals and an overall accuracy rate of 89.58% for
the EEG data considered
Classification of vertigo states based on The RF model had the best classification,
Yuehua and Jinxiang [141] Vertigo state
machine learning and EEG signal analysis with an accuracy rate of 82.5%
KNN classifier achieved good results in
Shuyi [142] Alcoholic Used NTFT and k-cross validation method
average accuracy which were up to 99%
The results obtained were 83% accurate in
Satyanarayana [143] Emotions An SVM emotion classifier based on EEG
detecting emotions
GD: Gender detection; FIR: Finite impulse response; RBF: Radial basis function; BCI: Brain–computer interface; SZ:
Schizophrenia; NTFT: Normal time–frequency transformation; HC: Healthy control; CR: Character recognition;
RBF: Radial basis function; ICH: Intracerebral hemorrhage; HRV: Heart rate variability; RESP: Respiratory rate;
DFA: Declined fluctuation analysis; (see other abbreviations noted in Table 4).

The advancement in EEG classification models has led to new possibilities in detecting
depression, a mental disorder that affects a significant portion of the global population.
To recognize depression based on EEG signals, some researchers have utilized the tree
model’s feature selection algorithm to establish a depression recognition model. It is clear
from Table 5 that research on the use of traditional ML algorithms to classify EEG signals is
still ongoing and is expanding. EEG signals are complex patterns of electrical activity in the
brain, and accurately classifying them can be crucial to understanding various neurological
conditions and cognitive processes.
Several research studies have utilized entropy measurement and statistical features of
EEG signals in gender detection to enhance its accuracy. To obtain EEG data on negative
and positive emotions for training and testing, a finite impulse response (FIR) filter model
is commonly employed. Decision trees, random forests, and multi-layer perceptron are
popularly used to predict gender from the obtained data. The findings suggest that the
random forest classifier performs best with the EEG of negative emotions, and investigates
the effect of excluding individual and multiple electrodes from the EEG data on the system
performance [136].
According to a study in [140], EEG can be used to detect neurophysiological changes
associated with schizophrenia. In another study, external vestibular electrical stimulation
was used to induce vertigo symptoms and EEG features were extracted using a wavelet
decomposition algorithm. The extracted features were then classified into different levels of
vertigo using logical regression, SVM, backpropagation, and RF classifiers. The RF model
demonstrated the highest accuracy of 82.5% [141]. Additionally, a regular time–frequency
transform technique was applied in [142] to predict EEG signals and evaluate individuals
with alcoholism at different stages. Furthermore, SVM was used for emotion recognition
based on EEG with large datasets [143].
Sensors 2023, 23, 6434 17 of 27

In addition to the previous points, EEG signals can also be used for fatigue detection.
In [144], an advanced machine learning method was proposed to use EEG signals to
detect driver fatigue and alert the driver as early as possible to prevent potential risks
while driving. This method is based on a flexible analytic wavelet transform. In [145],
they presented forehead EEG in combination with machine vision for detecting fatigue in
real-time. Experiments demonstrated that the proposed method could achieve significant
performance. In the field of aviation, in [146], they proposed using EEG to discriminate
aircraft pilot cognitive workload during flight, which achieved an accuracy of 91.67% in
classification tasks. Furthermore, in the maritime field, an approach was proposed for
assessing mental fatigue based on EEG frequency bands [147]. This approach was intended
for demanding maritime operations. The approach was tested in a realistic vessel simulator
and the results indicated that it could detect increased mental fatigue levels. EEG can also
be useful in the work context. For example, in [148], they proposed a measure that uses
implicit EEG signals to predict workers’ experience as a proxy for their ability to recognize
hazards. This leads to further improvement in the investigation of how we can derive
greater benefits from EEG signals.

2.5.2. Deep Learning

Although the structure of deep learning models is more complex compared to tradi-
tional ML, they offer greater advantages in classifying and predicting EEG signals. Several
researchers have utilized deep learning models to analyze EEG signals for disease detection
such as Alzheimer’s disease, epilepsy, ischemic stroke, etc. and to predict the progres-
sion of these diseases. A summary of these methods is discussed and reported in Table 6.
In Figure 4, we provide a clear view of how DL can be used for EEG data classification [149].
Initially, the EEG data are subjected to denoising and subsequent feature engineering. Af-
terwards, the processed data is converted into two-dimensional (2D) or three-dimensional
(3D) data, which serves as input to the CNN model. Finally, the CNN model is subjected to
training and optimization, and the optimal model is selected as the final model.

Table 6. Application of the deep learning method in EEG research.

Ref. Domain Proposed Method Conclusion

A method was proposed to generate a
suitable feature set using convolution and The method achieved 80% classification
Morabito [150] Alzheimer’s disease
then use full connectivity to make accuracy in Alzheimer’s disease
predictions
The system achieved nearly 90%
A deep learning processing system to reduce
Morabito [151] Alzheimer’s disease classification accuracy in diagnosing
the dimensionality of the feature space
Alzheimer’s disease
A novel end-to-end model designed for the
Their method achieved a high ROC-AUC
Kim [152] Alzheimer’s disease purpose of low-cost and noninvasive
score of 0.9
diagnosis of brain disorders
The method allowed for improved
A deep learning and multimodal fusion
diagnostic accuracy and earlier prediction
Kunekar [153] Epilepsy approach was proposed for the diagnosis of
of seizures due to the continuous
epilepsy
performance of the data
Proposed a simple CNN model to identify The CNN model achieved 98% accuracy in
Sagga [154] Epilepsy
epileptic seizures seizure detection
Using neural network model to process
The classification accuracy of
one-dimensional time series and
Qing [155] Epilepsy EfficientNetV2 model for epileptic EEG
two-dimensional EEG image EEG data types
was 98.69%
to detect seizures
A deep learning-based stroke evaluation CNN was 22.86% more accurate than
Ouyu [156] Ischemic stroke
model for stroke diagnosis logistic regression
Stroke detection using VGG-16 and The accuracy of the model in predicting
Kumar and Sengupta [157] Ischemic stroke
Resnet-50 models stroke reached 90%
Sensors 2023, 23, 6434 18 of 27

Table 6. Cont.

Ref. Domain Propose Method Conclusion

The accuracy of the results obtained in
A CNN DeprNet was proposed for recording split and subjective split
Seal [158] Depression
depression diagnosis experiments was 99.37% and 91.4%,
respectively
The accuracy reached 91.67% when all 19
Automatic detection of MDD Using EEG
Rafiei [159] Depression channels were used and 87.5% after the
data and deep neural network architecture
channels were reduced
Alexnet and GoogleNet used EEG signals to AlexNet was better at detecting sleep
Sudhakar [160] Sleep
detect sleep disorders disorders with an accuracy of 93.33%
Combined CNN and RNN to determine the
When considering all datasets, the highest
Leino [161] Sleep sleep stage of the EEG channel measured by
automatic scoring accuracy was 79.7%
AES
The optimized GoogleNet model was used
The accuracy of the sleep state of the EEG
Kang and Hong [162] Sleep to construct CNN automatic sleep stage
F4 channel was the highest at 77.6%
classification in single channel EEG
An end-to-end deep neural network could This model could successfully promote
Almogbel [163] Cognitive accommodate the original EEG signals from EEG signals and classify drivers’ cognitive
4 channels within a month as input workload with high accuracy
A highly accurate, EEG based driver fatigue Based on different indicators, the accuracy
Bhardwaj [164] Cognitive classification system to reduce fatigue of the deep learning automatic encoder
related road accidents was as high as 99.7%
CNN: Convolutional neural network; DL: Deep learning; MF: Multimodal fusion; MDD: Major depression
disorder; RNN: Recursive neural network; AES: Dynamic electrode set.

Figure 4. The pipeline of DL models for EEG classification [149]. The pipeline starts with feature
engineering applied to the EEG signal. The processed data is then transformed into 2D or 3D format,
which serves as input to a convolutional neural network (CNN) model. Finally, the CNN model
undergoes training and optimization, leading to the identification of an optimal model as the final
result of the pipeline.

Continuous development in stroke research has been achieved through the use of
deep learning models based on EEG signals. To prevent stroke effectively, a deep learning-
based stroke evaluation model has been used. This model extracts mel frequency cepstral
coefficient (MFCC) features and inputs them into a CNN, which achieves 22.86% higher
accuracy compared to logistic regression [156]. To extract more information from signals
passing through multiple convolution layers, hidden layers, and filters effectively, another
studies utilized VGG-16 and Resnet-50 models for stroke detection, resulting in a model
accuracy of 90% [157].
In order to detect MDD, EEG data has been analyzed using the DeprNet model [158,159].
Additionally, AlexNet [165] and GoogleNet [166] are utilized to identify sleep disturbances
from EEG signals through visual recognition tasks. Three categories of EEG signals were
analyzed, namely epilepsy, normal EEG, and sleep disorders. Results indicate that AlexNet
outperforms GoogleNet in detecting sleep disorders, achieving an accuracy of 93.33% [160].
In a separate study, a combination of CNN and RNN is employed to classify sleep stages
using the EEG channel Fp1/Fp2, achieving an accuracy of 79.7% [161]. The studies demon-
Sensors 2023, 23, 6434 19 of 27

strate the potential of deep learning algorithms in enhancing the automatic classification of
sleep stages based on EEG signals.
Using the EEG, a study investigated an end-to-end deep neural network for accurately
classifying drivers’ cognitive workload with high accuracy [163]. Additionally, a recent
study proposed a categorization system for driver fatigue that employs EEG signals in
conjunction with machine learning and deep learning algorithms. The results demonstrated
a significant level of precision in distinguishing between various fatigue states [164].
In [152], an approach aims to minimize human intervention while ensuring that all
the necessary components for EEG analysis are integrated in a logical and comprehensible
way. The model demonstrates significant performance in detecting Alzheimer’s disease
early, as indicated by its high ROC-AUC score of 0.9 [152].
The proposal utilized a deep learning method that employed EEG signals recorded by
the Muse EEG headband for performing emotion recognition tasks.
In [167], a proposal for an EEG-based brain–computer interface (BCI) was presented.
It uses a deep learning method that employed EEG signals recorded by the Muse EEG head-
band for performing emotion recognition tasks. Furthermore, in [168], a new lightweight
multidimensional attention network was proposed to address issues related to poor
generalization across datasets, high predicting volatility, and low model interpretability.
The method led to an enhanced classification performance in various BCI tasks.

3. Future Directions and Common Challenges

ML methods often face challenges related to data, as they may require larger datasets
than traditional methods to achieve similar performance. EEG data collection can be
complex and challenging, and continuous improvement of ML models is necessary to fit
older processing pipelines for better performance or to reduce the required amount of
data [169]. However, deep learning has simplified the EEG signal processing pipeline,
making it an end-to-end task [93]. Furthermore, deep learning has facilitated new research
avenues, such as generating images from EEG signals and transfer learning between
different fields [169–172]. Overall, EEG signals are a valuable source of information for
understanding brain activity, and both traditional and ML methods offer unique benefits
and challenges for processing these signals.
The matter of data heterogeneity in EEG research arises due to variations in the
acquisition devices employed across different datasets. Domain adaptation (DA) is a
technique employed to address the challenge of data heterogeneity by leveraging similar
domain data to reduce the data discrepancy [173]. The utilization of DA can serve as a
means of alleviating this particular issue. The authors of [174] introduced a multi-modal
domain adaptive variational autoencoder approach to enhance the performance of emotion
recognition tasks based on EEG data.
Additionally, the implementation of data privacy laws such as the General Data
Protection Regulation (GDPR) [175] can make it difficult to directly access source data for
training due to personally identifiable information from patients present in the EEG data.
Federated learning (FL) is a learning method that involves training multiple local models
and then obtaining a global model by aggregating these models globally without sharing
raw data [176]. In [177], the authors proposed a transferable FL technique to perform
EEG classification tasks. The experimental results show that the approach utilized by the
researchers can achieve an average accuracy of 91.10% using the Sleep-EDF dataset [178].
The field of EEG research offers access to several databases that contain EEG data.
These databases contain recordings of electrical brain activity obtained from multiple elec-
trodes and can be used to investigate a variety of research questions. For example, some
databases include EEG recordings of seizures from their onset to their end, while others
contain EEG data of participants performing different types of movements. Additionally,
some databases provide auditory-evoked EEG recordings, which capture the brain’s re-
sponse to sound stimuli. Researchers can use these datasets to advance their understanding
of brain function and behavior.
Sensors 2023, 23, 6434 20 of 27

As previously mentioned, EEG signal processing presents several challenges that are
summarized as follows.
• EEG data often contain noise and artifacts from various sources, such as muscle
movements, eye blinks, electrocardiogram signals, and electrical interference. These
unwanted components can significantly affect the quality of EEG signals.
• EEG signals are non-stationary, meaning that their statistical properties change over
time, making it difficult to analyze them using traditional methods. This characteristic
requires specialized techniques to capture the time-varying nature of EEG signals.
• EEG electrodes record signals originating from multiple sources in the brain, which
can result in a phenomenon called volume conduction. The superposition of signals
from multiple sources makes it challenging to locate the exact source of specific signals.
• The EEG signal acquisition measures the potential difference between the acting
electrode and the reference electrode. This leads to the problem of electrode reference.
The data obtained can vary depending on the selection of the reference electrode.
Selecting the best point for the reference electrode can be a challenging task.
• One of the challenges in EEG-based deep learning models is their interpretability.
If we can interpret the deep learning model accurately, patients may have more trust
in the machine learning diagnosis than in the diagnosis given by a doctor [89].
• EEG signals vary between individuals due to differences in skull thickness, con-
ductivity, and brain structure, making it difficult to compare data between subjects.
Specialized analysis methods must be employed to account for individual differences
while comparing EEG signals.
• Interpreting EEG signals requires expertise in both neuroscience and signal processing,
as they are indirect measures of neural activity. Proper analysis with different machine
learning algorithms might help to decode specific features of the signal that relate to
cognitive or behavioral states.
Addressing these challenges requires the development of new methods that can handle
these unique features of EEG signals, including denoising, source localization, improved
electrode configurations, and AI based on signal processing techniques.

4. Conclusions
This paper presents a comprehensive analysis of various techniques used for EEG
preprocessing and feature extraction. We also discuss EEG acquisition methods and sum-
marize signal denoising processes, including regression, blind source separation, wavelet
transform, and empirical mode decomposition. Our study focuses on time–frequency
analysis, high-order spectral analysis, and nonlinear dynamic analysis, and their applica-
tions in EEG feature engineering. We observed that machine learning algorithms have the
potential to achieve high accuracy in EEG classification, although the accuracy of classifiers
varies. We also found that deep learning models exhibit a comparable accuracy in detecting
seizures. To date, AI based algorithms have the potential to improve EEG analysis and
diagnosis, leading to better patient outcomes.

Author Contributions: Conceptualization, A.C.; methodology, A.C.; software, A.C.; validation, A.C.,
Y.W., R.K. and A.B.; formal analysis, A.C.; investigation, A.C. and Y.W.; resources, A.C.; data curation,
A.C.; writing—original draft preparation, A.C. and Y.W.; writing—review and editing, Y.W., R.K.
and A.B.; visualization, A.C.; supervision, A.C.; project administration, A.C.; funding acquisition,
A.C. All authors have read and agreed to the published version of the manuscript.
Funding: This research was funded by National Natural Science Foundation of China grant number
82260360, the Guilin Innovation Platform and Talent Program 20222C264164 and the Guangxi Science
and Technology Base and Talent Project (2022AC18004, 2022AC21040).
Institutional Review Board Statement: Not applicable.
Informed Consent Statement: Not applicable.
Data Availability Statement: Not applicable.
Sensors 2023, 23, 6434 21 of 27

Conflicts of Interest: The authors declare no conflict of interest. The funders had no role in the design
of the study.

References
1. Fan, J.; Fang, L.; Wu, J.; Guo, Y.; Dai, Q. From brain science to artificial intelligence. Engineering 2020, 6, 248–252. [CrossRef]
2. Lane, R.D.; Ryan, L.; Nadel, L.; Greenberg, L. Memory reconsolidation, emotional arousal, and the process of change in
psychotherapy: New insights from brain science. Behav. Brain Sci. 2015, 38, e1. [CrossRef]
3. da Silva, F.L. EEG and MEG: Relevance to neuroscience. Neuron 2013, 80, 1112–1128.
4. Da Silva, F.L. EEG: Origin and measurement. In EEG-fMRI: Physiological Basis, Technique, and Applications; Springer:
Berlin/Heidelberg, Germany, 2023; pp. 23–48.
5. Saibene, A.; Caglioni, M.; Corchs, S.; Gasparini, F. EEG-Based BCIs on Motor Imagery Paradigm Using Wearable Technologies: A
Systematic Review. Sensors 2023, 23, 2798. [CrossRef]
6. Islam, M.K.; Rastegarnia, A. Recent advances in EEG (non-invasive) based BCI applications. Front. Comput. Neurosci. 2023, 17,
1151852. [CrossRef]
7. Patil, A.U.; Lin, C.; Lee, S.H.; Huang, H.W.; Wu, S.C.; Madathil, D.; Huang, C.M. Review of EEG-based neurofeedback as a
therapeutic intervention to treat depression. Psychiatry Res. Neuroimaging 2023, 329, 111591. [CrossRef]
8. Mulert, C.; Lemieux, L. EEG-fMRI: Physiological Basis, Technique, and Applications; Springer: Berlin/Heidelberg, Germany, 2023.
9. Cecchetti, G.; Agosta, F.; Canu, E.; Basaia, S.; Barbieri, A.; Cardamone, R.; Bernasconi, M.P.; Castelnovo, V.; Cividini, C.; Cursi, M.;
et al. Cognitive, EEG, and MRI features of COVID-19 survivors: A 10-month study. J. Neurol. 2022, 269, 3400–3412. [CrossRef]
10. Chaddad, A. Brain function diagnosis enhanced using denoised fNIRS raw signals. J. Biomed. Sci. Eng. 2014, 2014, 43965.
[CrossRef]
11. Chaddad, A.; Kamrani, E.; Le Lan, J.; Sawan, M. Denoising fNIRS signals to enhance brain imaging diagnosis. In Proceedings of
the 2013 29th Southern Biomedical Engineering Conference, Miami, FL, USA, 3–5 May 2013; pp. 33–34.
12. Chaddad, A. Brain function evaluation using enhanced fNIRS signals extraction. In Proceedings of the 2014 48th Annual
Conference on Information Sciences and Systems (CISS), Princeton, NJ, USA, 19–21 March 2014; pp. 1–4.
13. Yoganathan, K.; Malek, N.; Torzillo, E.; Paranathala, M.; Greene, J. Neurological update: Structural and functional imaging in
epilepsy surgery. J. Neurol. 2023, 270, 2798–2808. [CrossRef]
14. Lopez-Gordo, M.A.; Sanchez-Morillo, D.; Valle, F.P. Dry EEG electrodes. Sensors 2014, 14, 12847–12870. [CrossRef]
15. Orban, M.; Elsamanty, M.; Guo, K.; Zhang, S.; Yang, H. A Review of Brain Activity and EEG-Based Brain–Computer Interfaces
for Rehabilitation Application. Bioengineering 2022, 9, 768. [CrossRef]
16. Liu, M.; Ushiba, J. Brain–machine Interface (BMI)-based Neurorehabilitation for Post-stroke Upper Limb Paralysis. Keio J. Med.
2022, 71, 82–92. [CrossRef]
17. Fischer, M.H.F.; Zibrandtsen, I.C.; Høgh, P.; Musaeus, C.S. Systematic Review of EEG Coherence in Alzheimer’s Disease. J.
Alzheimer’s Dis. 2023, 91, 1261–1272. [CrossRef]
18. Jiao, B.; Li, R.; Zhou, H.; Qing, K.; Liu, H.; Pan, H.; Lei, Y.; Fu, W.; Wang, X.; Xiao, X.; et al. Neural biomarker diagnosis and
prediction to mild cognitive impairment and Alzheimer’s disease using EEG technology. Alzheimer’s Res. Ther. 2023, 15, 1–14.
[CrossRef]
19. Shir, D.; Lazar, E.B.; Graff-Radford, J.; Aksamit, A.J.; Cutsforth-Gregory, J.K.; Jones, D.T.; Botha, H.; Ramanan, V.K.; Prusinski,
C.; Porter, A.; et al. Analysis of clinical features, diagnostic tests, and biomarkers in patients with suspected Creutzfeldt-Jakob
disease, 2014–2021. JAMA Netw. Open 2022, 5, e2225098. [CrossRef]
20. McClelland, V.M.; Fischer, P.; Foddai, E.; Dall’Orso, S.; Burdet, E.; Brown, P.; Lin, J.P. EEG measures of sensorimotor processing
and their development are abnormal in children with isolated dystonia and dystonic cerebral palsy. NeuroImage Clin. 2021,
30, 102569. [CrossRef]
21. Alvi, A.M.; Siuly, S.; Wang, H. A long short-term memory based framework for early detection of mild cognitive impairment
from EEG signals. IEEE Trans. Emerg. Top. Comput. Intell. 2022, 7, 375–388. [CrossRef]
22. Thakor, N.V.; Tong, S. Advances in quantitative electroencephalogram analysis methods. Annu. Rev. Biomed. Eng. 2004, 6, 453–495.
[CrossRef]
23. Lotte, F.; Bougrain, L.; Cichocki, A.; Clerc, M.; Congedo, M.; Rakotomamonjy, A.; Yger, F. A review of classification algorithms for
EEG-based brain–computer interfaces: A 10 year update. J. Neural Eng. 2018, 15, 031005. [CrossRef]
24. Aggarwal, S.; Chugh, N. Review of machine learning techniques for EEG based brain computer interface. Arch. Comput. Methods
Eng. 2022, 29, 3001–3020. [CrossRef]
25. Subha, D.P.; Joseph, P.K.; Acharya U, R.; Lim, C.M. EEG signal analysis: A survey. J. Med. Syst. 2010, 34, 195–212. [CrossRef]
26. Gaddipati, B.; Nelakuditi, U.R.; Medithe, J.W.C. Single lead EEG acquisition system for health care applications. In Proceedings
of the 2016 International Conference on Inventive Computation Technologies (ICICT), Coimbatore, India, 26–27 August 2016;
Volume 1, pp. 1–4.
27. Choi, H.; Park, J.; Yang, Y.M. A Novel Quick-Response Eigenface Analysis Scheme for Brain–Computer Interfaces. Sensors 2022,
22, 5860. [CrossRef]
28. Haas, L.F. Hans berger (1873–1941), richard caton (1842–1926), and electroencephalography. J. Neurol. Neurosurg. Psychiatry 2003,
74, 9. [CrossRef]
Sensors 2023, 23, 6434 22 of 27

29. Stytsenko, K.; Jablonskis, E.; Prahm, C. Evaluation of consumer EEG device Emotiv EPOC. In Proceedings of the MEi: CogSci
Conference, Ljubljana, Slovenia, 21–23 September 2011; p. 99.
30. Holewa, K.; Nawrocka, A. Emotiv EPOC neuroheadset in brain–computer interface. In Proceedings of the 2014 15th International
Carpathian Control Conference (ICCC), Velke Karlovice, Czech Republic, 28–30 May 2014; pp. 149–152.
31. Duvinage, M.; Castermans, T.; Dutoit, T.; Petieau, M.; Hoellinger, T.; Saedeleer, C.D.; Seetharaman, K.; Cheron, G. A P300-based
quantitative comparison between the Emotiv Epoc headset and a medical EEG device. Biomed. Eng. 2012, 765, 2012–2764.
32. Martins, R.; Selberherr, S.; Vaz, F. A CMOS IC for portable EEG acquisition systems. IEEE Trans. Instrum. Meas. 1998, 47, 1191–1196.
[CrossRef]
33. Cui, Y.; Tian, F.; Zhao, Q.; Hu, B. Design and Application of a Portable Sleep Inertia Detection System Based on EEG Signals.
In Proceedings of the 2021 IEEE International Conference on Bioinformatics and Biomedicine (BIBM), Houston, TX, USA, 9–12
December 2021; pp. 3012–3017.
34. Yu, Y.; Li, N.; Li, Y.; Liu, W. A portable waterproof EEG acquisition device for dolphins. Sensors 2021, 21, 3336. [CrossRef]
35. Gottlibe, M.; Rosen, O.; Weller, B.; Mahagney, A.; Omar, N.; Khuri, A.; Srugo, I.; Genizi, J. Stroke identification using a portable
EEG device–A pilot study. Neurophysiol. Clin. 2020, 50, 21–25. [CrossRef]
36. Chen, J.; Li, X.; Mi, X.; Pan, S. A high precision EEG acquisition system based on the CompactPCI platform. In Proceedings of the
2014 7th International Conference on Biomedical Engineering and Informatics, Dalian, China, 14–16 October 2014; pp. 511–516.
37. Swann, N.; Poizner, H.; Houser, M.; Gould, S.; Greenhouse, I.; Cai, W.; Strunk, J.; George, J.; Aron, A.R. Deep brain stimulation
of the subthalamic nucleus alters the cortical profile of response inhibition in the beta frequency band: A scalp EEG study in
Parkinson’s disease. J. Neurosci. 2011, 31, 5721–5729. [CrossRef]
38. Li, B.; Cheng, T.; Guo, Z. A review of EEG acquisition, processing and application. J. Phys. Conf. Ser. 2021, 1907, 012045.
[CrossRef]
39. Bajaj, N.; Carrión, J.R.; Bellotti, F. Phyaat: Physiology of auditory attention to speech dataset. arXiv 2020, arXiv:2005.11577.
40. Luck, S.J. An Introduction to the Event-Related Potential Technique; MIT Press: Cambridge, MA USA, 2014.
41. Wang, J.; Wang, M. Review of the emotional feature extraction and classification using EEG signals. Cogn. Robot. 2021, 1, 29–40.
[CrossRef]
42. Ma, C.; Zhang, M.; Sun, X.; Wang, H.; Gao, Z. Dynamic threshold distribution domain adaptation network: A cross-subject
fatigue recognition method based on EEG signals. IEEE Trans. Cogn. Dev. Syst. 2023, early access. [CrossRef]
43. Gamage, T.A.; Kalansooriya, L.P.; Sandamali, E.R.C. An Emotion Classification Model for Driver Emotion Recognition Using
Electroencephalography (EEG). In Proceedings of the 2022 International Research Conference on Smart Computing and Systems
Engineering (SCSE), Colombo, Sri Lanka, 1 September 2022; Volume 5, pp. 76–82.
44. Shen, J.; Zhang, Y.; Liang, H.; Zhao, Z.; Dong, Q.; Qian, K.; Zhang, X.; Hu, B. Exploring the Intrinsic Features of EEG Signals via
Empirical Mode Decomposition for Depression Recognition. IEEE Trans. Neural Syst. Rehabil. Eng. 2023, 31, 356–365. [CrossRef]
45. Saedi, S.; Fini, A.A.F.; Khanzadi, M.; Wong, J.; Sheikhkhoshkar, M.; Banaei, M. Applications of electroencephalography in
construction. Autom. Constr. 2022, 133, 103985. [CrossRef]
46. Han, C.H.; Choi, G.Y.; Hwang, H.J. Deep Convolutional Neural Network Based Eye States Classification Using Ear-EEG. Expert
Syst. Appl. 2022, 192, 116443. [CrossRef]
47. Pawuś, D.; Paszkiel, S. The application of integration of EEG signals for authorial classification algorithms in implementation for
a mobile robot control using movement imagery—Pilot study. Appl. Sci. 2022, 12, 2161. [CrossRef]
48. Chen, X.; Li, C.; Liu, A.; McKeown, M.J.; Qian, R.; Wang, Z.J. Toward open-world electroencephalogram decoding via deep
learning: A comprehensive survey. IEEE Signal Process. Mag. 2022, 39, 117–134. [CrossRef]
49. Pei, W.; Wu, X.; Zhang, X.; Zha, A.; Tian, S.; Wang, Y.; Gao, X. A pre-gelled EEG electrode and its application in SSVEP-based BCI.
IEEE Trans. Neural Syst. Rehabil. Eng. 2022, 30, 843–850. [CrossRef]
50. Jemal, I.; Mezghani, N.; Abou-Abbas, L.; Mitiche, A. An interpretable deep learning classifier for epileptic seizure prediction
using EEG data. IEEE Access 2022, 10, 60141–60150. [CrossRef]
51. Wen, D.; Li, R.; Jiang, M.; Li, J.; Liu, Y.; Dong, X.; Saripan, M.I.; Song, H.; Han, W.; Zhou, Y. Multi-dimensional conditional mutual
information with application on the EEG signal analysis for spatial cognitive ability evaluation. Neural Netw. 2022, 148, 23–36.
[CrossRef]
52. Li, G.; Ouyang, D.; Yuan, Y.; Li, W.; Guo, Z.; Qu, X.; Green, P. An EEG data processing approach for emotion recognition. IEEE
Sens. J. 2022, 22, 10751–10763. [CrossRef]
53. Freismuth, D.; TaheriNejad, N. On the treatment and diagnosis of attention deficit hyperactivity disorder with eeg assistance.
Electronics 2022, 11, 606. [CrossRef]
54. Jiang, X.; Bian, G.B.; Tian, Z. Removal of artifacts from EEG signals: A review. Sensors 2019, 19, 987. [CrossRef] [PubMed]
55. Sheoran, M.; Kumar, S.; Chawla, S. Methods of denoising of electroencephalogram signal: A review. Int. J. Biomed. Eng. Technol.
2015, 18, 385–395. [CrossRef]
56. Kenemans, J.L.; Molenaar, P.C.; Verbaten, M.N.; Slangen, J.L. Removal of the ocular artifact from the EEG: A comparison of time
and frequency domain methods with simulated and real data. Psychophysiology 1991, 28, 114–121. [CrossRef] [PubMed]
57. Croft, R.J.; Barry, R.J. Removal of ocular artifact from the EEG: A review. Neurophysiol. Clin. Neurophysiol. 2000, 30, 5–19.
[CrossRef]
Sensors 2023, 23, 6434 23 of 27

58. Ranjan, R.; Sahana, B.C.; Bhandari, A.K. Ocular artifact elimination from electroencephalography signals: A systematic review.
Biocybern. Biomed. Eng. 2021, 41, 960–996. [CrossRef]
59. Stachaczyk, M.; Atashzar, S.F.; Farina, D. Adaptive spatial filtering of high-density EMG for reducing the influence of noise and
artefacts in myoelectric control. IEEE Trans. Neural Syst. Rehabil. Eng. 2020, 28, 1511–1517. [CrossRef]
60. Kim, H.C.; Lee, J.H. Systematic evaluation of recursive approach of EEG-segment-based PCA for removal of helium-pump
artefact from MRI. Electron. Lett. 2022, 58, 567–569. [CrossRef]
61. Noorbasha, S.K.; Sudha, G.F. Removal of EOG artifacts and separation of different cerebral activity components from single
channel EEG—An efficient approach combining SSA–ICA with wavelet thresholding for BCI applications. Biomed. Signal Process.
Control 2021, 63, 102168. [CrossRef]
62. Stergiadis, C.; Kostaridou, V.D.; Klados, M.A. Which BSS method separates better the EEG Signals? A comparison of five different
algorithms. Biomed. Signal Process. Control 2022, 72, 103292. [CrossRef]
63. Mijović, B.; De Vos, M.; Gligorijević, I.; Taelman, J.; Van Huffel, S. Source Separation From Single-Channel Recordings by Combining
Empirical-Mode Decomposition and Independent Component Analysis. IEEE Trans. Biomed. Eng. 2010, 57, 2188–2196. [CrossRef]
64. Onton, J.; Westerfield, M.; Townsend, J.; Makeig, S. Imaging human EEG dynamics using independent component analysis.
Neurosci. Biobehav. Rev. 2006, 30, 808–822. [CrossRef] [PubMed]
65. Li, Y.; Wang, P.T.; Vaidya, M.P.; Flint, R.D.; Liu, C.Y.; Slutzky, M.W.; Do, A.H. Electromyogram (EMG) Removal by Adding
Sources of EMG (ERASE)—A novel ICA-based algorithm for removing myoelectric artifacts from EEG. Front. Neurosci. 2021,
14, 597941. [CrossRef] [PubMed]
66. Maddirala, A.K.; Veluvolu, K.C. ICA With CWT and k-means for Eye-Blink Artifact Removal From Fewer Channel EEG. IEEE
Trans. Neural Syst. Rehabil. Eng. 2022, 30, 1361–1373. [CrossRef]
67. Patel, R.; Sengottuvel, S.; Janawadkar, M.; Gireesan, K.; Radhakrishnan, T.; Mariyappa, N. Ocular artifact suppression from
EEG using ensemble empirical mode decomposition with principal component analysis. Comput. Electr. Eng. 2016, 54, 78–86.
[CrossRef]
68. Xie, Z.; Yu, B.; Xu, W. A Learning Model of Evoked EEG Signals Based on PCA and Semi-supervised SVM. In Proceedings of the
2021 4th International Conference on Robotics, Control and Automation Engineering (RCAE), Wuhan, China, 4–6 November
2021; pp. 121–127.
69. Sheoran, P.; Saini, J. A new method for automatic electrooculogram and eye blink artifacts correction of EEG signals using CCA
and NAPCT. Procedia Comput. Sci. 2020, 167, 1761–1770. [CrossRef]
70. Mert, A.; Akan, A. Detrended fluctuation thresholding for empirical mode decomposition based denoising. Digit. Signal Process.
2014, 32, 48–56. [CrossRef]
71. De Clercq, W.; Vergult, A.; Vanrumste, B.; Van Paesschen, W.; Van Huffel, S. Canonical correlation analysis applied to remove
muscle artifacts from the electroencephalogram. IEEE Trans. Biomed. Eng. 2006, 53, 2583–2587. [CrossRef]
72. Hassan, M.; Boudaoud, S.; Terrien, J.; Karlsson, B.; Marque, C. Combination of canonical correlation analysis and empirical mode
decomposition applied to denoising the labor electrohysterogram. IEEE Trans. Biomed. Eng. 2011, 58, 2441–2447. [CrossRef]
73. Miao, M.; Hu, W.; Xu, B.; Zhang, J.; Rodrigues, J.J.P.C.; de Albuquerque, V.H.C. Automated CCA-MWF Algorithm for
Unsupervised Identification and Removal of EOG Artifacts From EEG. IEEE J. Biomed. Health Inform. 2022, 26, 3607–3617.
[CrossRef] [PubMed]
74. Ranjan, R.; Sahana, B.C.; Bhandari, A.K. Cardiac Artifact Noise Removal From Sleep EEG Signals Using Hybrid Denoising Model.
IEEE Trans. Instrum. Meas. 2022, 71, 1–10. [CrossRef]
75. Yan, X.; Boudrias, M.H.; Mitsis, G.D. Removal of Transcranial Alternating Current Stimulation EEG Artifacts Using Blind Source
Separation and Wavelets. IEEE Trans. Biomed. Eng. 2022, 69, 3183–3192. [CrossRef] [PubMed]
76. Mowla, M.R.; Ng, S.C.; Zilany, M.S.; Paramesran, R. Artifacts-matched blind source separation and wavelet transform for
multichannel EEG denoising. Biomed. Signal Process. Control 2015, 22, 111–118. [CrossRef]
77. Dautov, C.P.; Ozerdem, M.S. Wavelet transform and signal denoising using Wavelet method. In Proceedings of the 2018 26th
Signal Processing and Communications Applications Conference (SIU), Izmir, Turkey, 2–5 May 2018; pp. 1–4.
78. Khatun, S.; Mahajan, R.; Morshed, B.I. Comparative study of wavelet-based unsupervised ocular artifact removal techniques for
single-channel EEG data. IEEE J. Transl. Eng. Health Med. 2016, 4, 1–8. [CrossRef]
79. Aqil, M.; Jbari, A.; Bourouhou, A. ECG Signal Denoising by Discrete Wavelet Transform. Int. J. Online Eng. 2017, 13, 51–68.
[CrossRef]
80. Zhou, W.; Gotman, J. Removal of EMG and ECG artifacts from EEG based on wavelet transform and ICA. In Proceedings of the
The 26th Annual International Conference of the IEEE Engineering in Medicine and Biology Society, San Francisco, CA, USA, 1–5
September 2004; Volume 1, pp. 392–395.
81. Tibdewal, M.N.; Mahadevappa, M.; Ray, A.K.; Malokar, M.; Dey, H.R. Power line and ocular artifact denoising from EEG using
notch filter and wavelet transform. In Proceedings of the 2016 3rd International Conference on Computing for Sustainable Global
Development (INDIACom), New Delhi, India, 16–18 March 2016; pp. 1654–1659.
82. Zhang, D.X.; Wu, X.P.; Guo, X.J. The EEG Signal Preprocessing Based on Empirical Mode Decomposition. In Proceedings
of the 2008 2nd International Conference on Bioinformatics and Biomedical Engineering, Shanghai, China, 16–18 May 2008;
pp. 2131–2134.
Sensors 2023, 23, 6434 24 of 27

83. Chen, X.; Chen, Q.; Zhang, Y.; Wang, Z.J. A Novel EEMD-CCA Approach to Removing Muscle Artifacts for Pervasive EEG. IEEE
Sens. J. 2019, 19, 8420–8431. [CrossRef]
84. Yang, H.W.; Jeng, S.K.; Young, H.W.V.; Lin, C.; Wang, Y.H.; Hu, K.; Lo, M.T. A Minimum Arclength Method for Removing Spikes
in Empirical Mode Decomposition. IEEE Access 2019, 7, 13284–13294. [CrossRef]
85. Li, M.; Zhang, Y. An improved MAMA-EMD for the automatic removal of EOG artifacts. Biocybern. Biomed. Eng. 2021,
41, 1182–1196. [CrossRef]
86. Alyasseri, Z.A.A.; Khader, A.T.; Al-Betar, M.A.; Abasi, A.K.; Makhadmeh, S.N. EEG signals denoising using optimal wavelet
transform hybridized with efficient metaheuristic methods. IEEE Access 2019, 8, 10584–10605. [CrossRef]
87. Chaddad, A.; Peng, J.; Xu, J.; Bouridane, A. Survey of Explainable AI Techniques in Healthcare. Sensors 2023, 23, 634. [CrossRef]
88. Smilkov, D.; Thorat, N.; Kim, B.; Viégas, F.; Wattenberg, M. Smoothgrad: Removing noise by adding noise. arXiv 2017,
arXiv:1706.03825.
89. Torres, J.M.M.; Medina-DeVilliers, S.; Clarkson, T.; Lerner, M.D.; Riccardi, G. Evaluation of interpretability for deep learning
algorithms in EEG emotion recognition: A case study in autism. Artif. Intell. Med. 2023, 143, 102545. [CrossRef]
90. Mansour, M.; Khnaisser, F.; Partamian, H. An explainable model for eeg seizure detection based on connectivity features. arXiv
2020, arXiv:2009.12566.
91. Sturm, I.; Lapuschkin, S.; Samek, W.; Muller, K.R. Interpretable deep neural networks for single-trial EEG classification. J.
Neurosci. Methods 2016, 274, 141–145. [CrossRef] [PubMed]
92. Hartmann, K.G.; Schirrmeister, R.T.; Ball, T. EEG-GAN: Generative adversarial networks for electroencephalograhic (EEG) brain
signals. arXiv 2018, arXiv:1806.01875.
93. Roy, Y.; Banville, H.; Albuquerque, I.; Gramfort, A.; Falk, T.H.; Faubert, J. Deep learning-based electroencephalography analysis:
A systematic review. J. Neural Eng. 2019, 16, 051001. [CrossRef]
94. Boashash, B.; Carson, H.; Mesbah, M. Detection of seizures in newborns using time-frequency analysis of EEG signals. In
Proceedings of the Tenth IEEE Workshop on Statistical Signal and Array Processing (Cat. No. 00TH8496), Pocono Manor, PA,
USA, 16 August 2000; pp. 564–568.
95. Hassanpour, H.; Mesbah, M.; Boashash, B. EEG spike detection using time-frequency signal analysis. In Proceedings of the 2004
IEEE International Conference on Acoustics, Speech, and Signal Processing, Montreal, QC, Canada, 17–21 May 2004; Volume 5,
pp.V–V.
96. Khan, N.A.; Mohammadi, M.; Ghafoor, M.; Tariq, S.A. Convolutional Neural Networks Based Time-Frequency Image Enhance-
ment For the Analysis of EEG Signals. Multidimens. Syst. Signal Process. 2022, 33, 863–877. [CrossRef]
97. Mahmoodian, N.; Haddadnia, J.; Illanes, A.; Boese, A.; Friebe, M. Seizure prediction with cross-higher-order spectral analysis of
EEG signals. Signal Image Video Process. 2020, 14, 821–828. [CrossRef]
98. Murariu, M.G.; Tărniceriu, D.; Hris, că-Eva, O.D.; Lazăr, A.M. An Approach to Identify Different Types of EEG Epileptic Signals
Based on Higher-Order Spectra (HOS) Features. In Proceedings of the 2022 14th International Conference on Electronics,
Computers and Artificial Intelligence (ECAI), Ploiesti, Romania, 30 June–1 July 2022; pp. 1–5.
99. Du, X.; Dua, S.; Acharya, R.U.; Chua, C.K. Classification of epilepsy using high-order spectra features and principle component
analysis. J. Med. Syst. 2012, 36, 1731–1743. [CrossRef]
100. Hosseini, S.A.; Houshmand, M. Analysis of the EEG Signal Using Higher-Order Spectra (HOS) in the Neuro-marketing
Application. New Mark. Res. J. 2022, 12, 25–42.
101. Li, S.; Cha, S.H. Feature extraction based on high order statistics measures and entropy for eeg biometrics. In Proceedings of the
2019 7th International Workshop on Biometrics and Forensics (IWBF), Cancun, Mexico, 2–3 May 2019; pp. 1–6.
102. Xu, C.; Wang, H.; Huang, J. The Analysis of EEG Signals in Driving Behavior Based on Nonlinear Dynamics. In Proceedings
of the 2022 International Symposium on Control Engineering and Robotics (ISCER), Changsha, China, 18–20 February 2022;
pp. 271–276.
103. Shuchun, L.; Diankui, L.; Xiaofeng, Z.; Xiaomin, L.; Yang, L. Research on nonlinear dynamics of high-frequency EEG based on
correlation dimension and Lyapunov exponent. Mod. Biomed. Prog. 2017, 17, 561–566.
104. Dongmei, L. Research on Classification, Location and Prediction Methods of Epileptic EEG Signals Based on Nonlinear Dynamic
Characteristics. Master’s Thesis, Xinjiang Medical University, Ürümqi, China, 2017.
105. Dongwei, C.; Junjie, C. Research on the complexity of Lempel Ziv EEG signals in emotion recognition. J. Taiyuan Univ. Technol.
2014, 45, 758–763.
106. Meng, C.; Ning, Z.; Youjun, L.; Yan, L.; Qiang, H. Emotional EEG signal analysis based on equal symbolic entropy. Comput. Appl.
Res. 2018, 35, 2051–2054. 2091.
107. Anuragi, A.; Sisodia, D.S.; Pachori, R.B. EEG-based cross-subject emotion recognition using Fourier-Bessel series expansion based
empirical wavelet transform and NCA feature selection method. Inf. Sci. 2022, 610, 508–524. [CrossRef]
108. Sharma, N.; Pachori, R.B. Analysis of EEG signals using nonlinear dynamics: A review. Biocybern. Biomed. Eng. 2021, 41, 716–742.
109. Hazarika, B.B.; Gupta, D.; Kumar, B. EEG Signal Classification Using a Novel Universum-Based Twin Parametric-Margin Support
Vector Machine. Cogn. Comput. 2023, 1–16. [CrossRef]
110. Wang, J.; Gao, R.; Zheng, H.; Zhu, H.; Shi, C.J.R. SSGCNet: A Sparse Spectra Graph Convolutional Network for Epileptic EEG
Signal Classification. IEEE Trans. Neural Netw. Learn. Syst. 2023, early access. [CrossRef]
Sensors 2023, 23, 6434 25 of 27

111. Hassan, F.; Hussain, S.F.; Qaisar, S.M. Fusion of multivariate EEG signals for schizophrenia detection using CNN and machine
learning techniques. Inf. Fusion 2023, 92, 466–478. [CrossRef]
112. Zhang, X.L.; Begleiter, H.; Porjesz, B.; Wang, W.; Litke, A. Event related potentials during object recognition tasks. Brain Res. Bull.
1995, 38, 531–538. [CrossRef]
113. Koelstra, S.; Muhl, C.; Soleymani, M.; Lee, J.S.; Yazdani, A.; Ebrahimi, T.; Pun, T.; Nijholt, A.; Patras, I. Deap: A database for
emotion analysis; using physiological signals. IEEE Trans. Affect. Comput. 2011, 3, 18–31. [CrossRef]
114. Zheng, W.L.; Lu, B.L. Investigating critical frequency bands and channels for EEG-based emotion recognition with deep neural
networks. IEEE Trans. Auton. Ment. Dev. 2015, 7, 162–175. [CrossRef]
115. Ang, K.K.; Chin, Z.Y.; Wang, C.; Guan, C.; Zhang, H. Filter bank common spatial pattern algorithm on BCI competition IV
datasets 2a and 2b. Front. Neurosci. 2012, 6, 39. [CrossRef] [PubMed]
116. Tangermann, M.; Müller, K.R.; Aertsen, A.; Birbaumer, N.; Braun, C.; Brunner, C.; Leeb, R.; Mehring, C.; Miller, K.J.; Mueller-Putz,
G.; et al. Review of the BCI competition IV. Front. Neurosci. 2012, 6, 55. [CrossRef] [PubMed]
117. Sajda, P.; Gerson, A.; Muller, K.R.; Blankertz, B.; Parra, L. A data analysis competition to evaluate machine learning algorithms
for use in brain–computer interfaces. IEEE Trans. Neural Syst. Rehabil. Eng. 2003, 11, 184–185. [CrossRef]
118. Andrzejak, R.G.; Lehnertz, K.; Mormann, F.; Rieke, C.; David, P.; Elger, C.E. Indications of nonlinear deterministic and finite-
dimensional structures in time series of brain electrical activity: Dependence on recording region and brain state. Phys. Rev. E
2001, 64, 061907. [CrossRef]
119. Shoeb, A.H. Application of Machine Learning to Epileptic Seizure Onset Detection and Treatment. Ph.D. Thesis, Massachusetts
Institute of Technology, Cambridge, MA, USA, 2009.
120. Detti, P.; Vatti, G.; Zabalo Manrique de Lara, G. Eeg synchronization analysis for seizure prediction: A study on data of
noninvasive recordings. Processes 2020, 8, 846. [CrossRef]
121. Zhang, J.; Yin, Z.; Wang, R. Recognition of mental workload levels under complex human–machine collaboration by using
physiological features and adaptive support vector machines. IEEE Trans. Hum.-Mach. Syst. 2014, 45, 200–214. [CrossRef]
122. Venkatachalam, K.; Devipriya, A.; Maniraj, J.; Sivaram, M.; Ambikapathy, A.; Iraj, S.A. A Novel Method of motor imagery
classification using eeg signal. Artif. Intell. Med. 2020, 103, 101787.
123. Zhang, H.; Zhao, M.; Wei, C.; Mantini, D.; Li, Z.; Liu, Q. EEGdenoiseNet: A benchmark dataset for end-to-end deep learning
solutions of EEG denoising. arXiv 2020, arXiv:2009.11662.
124. Hosseini, M.P.; Hosseini, A.; Ahi, K. A review on machine learning for EEG signal processing in bioengineering. IEEE Rev.
Biomed. Eng. 2020, 14, 204–218. [CrossRef] [PubMed]
125. Choubey, H.; Pandey, A. A combination of statistical parameters for the detection of epilepsy and EEG classification using ANN
and KNN classifier. Signal Image Video Process. 2021, 15, 475–483.
126. Ping, L.; Yan, Z.; Jun, W. Support vector classifier for EEG signals based on nonlinear feature extraction. J. Shantou Univ. Nat. Sci.
Ed. 2009, 24, 6.
127. Jamunadevi, C.; Ragupathy, P.; Sritha, P.; Pandikumar, S.; Deepa, S. Performance Analysis of Random Forest Classifier in
Extracting Features from the EEG signal. In Proceedings of the 2022 International Conference on Advanced Computing
Technologies and Applications (ICACTA), Coimbatore, India, 4–5 March 2022; pp. 1–5.
128. Jiahui, C.; Youguo, W.; Qiqing, Z. Noise benefit in motion imagination classification based on K-nearest neighbor. Comput.
Technol. Dev. 2022, 32, 79–84.
129. Jiaying, L.; Li, Z.; Yan, B.; Fangqing, G. Research on classification of lower limb motor imagery EEG signals based on LDA and
KNN. Foreign Electron. Meas. Technol. 2021, 1.
130. Dongare, S.; Padole, D. Categorization of EEG Using Hybrid Features and Voting classifier for Motor Imagination. In Proceedings
of the 2021 International Conference on Recent Trends on Electronics, Information, Communication & Technology (RTEICT),
Bangalore, India, 27–28 August 2021; pp. 217–220.
131. Ren, W.; di, B.; Shuqi, W.; Xuemeng, L.; Dongsheng, Z. The application of decision tree algorithm in the prediction of stroke risk
classification. Chin. Conval. Med. 2019, 28, 4.
132. Huaiwen, C.; Yin, Y. Building a risk prediction model for ischemic stroke in Jiangxi based on machine learning. West. Med. 2022,
34, 1182-1187.
133. Hanqi, C.; Hao, Z.; Xiaomin, G.; Mingyang, P.; Guanghui, X.; Guozhong, C.; Xindao, Y.; Yu, X. Prediction of prognosis of
mechanical thrombectomy in acute stroke by machine learning combined with imaging features. J. Nanjing Med. Univ. Nat. Sci.
Ed. 2022, 42.
134. Yong, S.; Liqiang, W.; Fen, W.; Guoqiang, C.; Yingchao, Z.; Yahui, L.; Ling, Q.; Guangming, Z. Explainable machine learning
model is used to predict long-term cerebral ischemic events. Prev. Treat. Cardiovasc. Cerebrovasc. Dis. 2022, 22, 5.
135. Carrara, I.; Papadopoulo, T. Classification of BCI-EEG based on augmented covariance matrix. arXiv 2023, arXiv:2302.04508.
136. Alharbi, Y.F.; Alotaibi, Y.A. The Correlate of Emotion and Gender Classification Using EEG Signals. In Proceedings of the 2021
IEEE 6th International Conference on Signal and Image Processing (ICSIP), Nanjing, China, 22–24 October 2021; pp. 790–794.
137. Parmar, S.K.; Ramwala, O.A.; Paunwala, C.N. Performance Evaluation of SVM with Non-Linear Kernels for EEG-based Dyslexia
Detection. In Proceedings of the 2021 IEEE 9th Region 10 Humanitarian Technology Conference (R10-HTC), Bangalore, India, 30
September–2 October 2021; pp. 1–6.
138. Ling, H.; Aihua, Z. Application of improved decision tree SVM in EEG recognition. Comput. Eng. Des. 2010, 2.
Sensors 2023, 23, 6434 26 of 27

139. Hossain, M.S.; Hasan, T.; Hasan, M.M.; Rahman, M.M.; Sabiha, M.M. English Character recognition using EEG-based Visual
stimulations: A Machine Learning Analysis. In Proceedings of the 2022 International Conference on Innovations in Science,
Engineering and Technology (ICISET), Chittagong, Bangladesh, 26–27 February 2022; pp. 373–378.
140. Padayatty, R.V.; K, T.F.N. Detection of schizophrenia using EEG signals: A Machine learning approach. In Proceedings of the
2022 International Conference on Futuristic Technologies in Control Systems & Renewable Energy (ICFCR), Malappuram, India,
21–22 July 2022; pp. 1–8.
141. Yuehua, G.; Jinxiang, S. Classification of vertigo states combined with machine learning and EEG signal analysis. China Tissue
Eng. Res. 2022, 26.
142. Shuyi, Z.; Xiaoyan, L.; Jiansong, Z.; Gang, S. Eeg signal analysis method based on standard time-frequency transform. J. Electron.
Meas. Instrum. 2018, 7.
143. Satyanarayana, K.N.V.; Shankar, T.; Poojita, G.; Vinay, G.; Amaranadh, H.N.S.V.l.S.; Babu, A.G. An Approach to EEG based
Emotion Identification by SVM classifier. In Proceedings of the 2022 6th International Conference on Computing Methodologies
and Communication (ICCMC), Erode, India, 29–31 March 2022; pp. 650–654. [CrossRef]
144. Subasi, A.; Saikia, A.; Bagedo, K.; Singh, A.; Hazarika, A. EEG-based driver fatigue detection using FAWT and multiboosting
approaches. IEEE Trans. Ind. Inform. 2022, 18, 6602–6609. [CrossRef]
145. Min, J.; Cai, M.; Gou, C.; Xiong, C.; Yao, X. Fusion of forehead EEG with machine vision for real-time fatigue detection in an
automatic processing pipeline. Neural Comput. Appl. 2023, 35, 8859–8872. [CrossRef]
146. Taheri Gorji, H.; Wilson, N.; VanBree, J.; Hoffmann, B.; Petros, T.; Tavakolian, K. Using machine learning methods and EEG to
discriminate aircraft pilot cognitive workload during flight. Sci. Rep. 2023, 13, 2507. [CrossRef]
147. Monteiro, T.G.; Skourup, C.; Zhang, H. A task agnostic mental fatigue assessment approach based on EEG frequency bands for
demanding maritime operation. IEEE Instrum. Meas. Mag. 2021, 24, 82–88. [CrossRef]
148. Wang, M.; Zhao, Y.; Liao, P.C. EEG-based work experience prediction using hazard recognition. Autom. Constr. 2022, 136, 104151.
[CrossRef]
149. Bird, J.J.; Faria, D.R.; Manso, L.J.; Ayrosa, P.P.; Ekart, A. A study on CNN image classification of EEG signals represented in 2D
and 3D. J. Neural Eng. 2021, 18, 026005. [CrossRef]
150. Morabito, F.C.; Campolo, M.; Ieracitano, C.; Ebadi, J.M.; Bonanno, L.; Bramanti, A.; Desalvo, S.; Mammone, N.; Bramanti, P. Deep
convolutional neural networks for classification of mild cognitive impaired and Alzheimer’s disease patients from scalp EEG
recordings. In Proceedings of the 2016 IEEE 2nd International Forum on Research and Technologies for Society and Industry
Leveraging a better tomorrow (RTSI), Bologna, Italy, 7–9 September 2016; pp. 1–6.
151. Morabito, F.; Campolo, M.; Mammone, N.; Mario, V.; Silvana, F.; Tagliavini, F.; Sofia, V.; Fatuzzo, D.; Gambardella, A.; Labate, A.;
et al. Deep Learning Representation from Electroencephalography of Early-Stage Creutzfeldt-Jakob Disease and Features for
Differentiation from Rapidly Progressive Dementia. Int. J. Neural Syst. 2016, 27, 1650039. [CrossRef]
152. Kim, M.j.; Youn, Y.C.; Paik, J. Deep learning-based EEG analysis to classify normal, mild cognitive impairment, and dementia:
Algorithms and dataset. NeuroImage 2023, 272, 120054. [CrossRef] [PubMed]
153. Kunekar, P.R.; Gupta, M.; Agarwal, B. Deep Learning with Multi Modal Ensemble Fusion for Epilepsy Diagnosis. In Proceedings
of the 2020 3rd International Conference on Emerging Technologies in Computer Engineering: Machine Learning and Internet of
Things (ICETCE), Jaipur, India, 7–8 February 2020; pp. 80–84.
154. Sagga, D.; Echtioui, A.; Khemakhem, R.; Kallel, F.; Hamida, A.B. Epileptic Seizures Detection on EEG Signal Using Deep Learning
Techniques. In Proceedings of the 2022 6th International Conference on Advanced Technologies for Signal and Image Processing
(ATSIP), Sfax, Tunisia, 24–27 May 2022; pp. 1–6.
155. Qing, X.; Cheng, G.; Biao, C.; Shan, C. Classification of epileptic EEG signals based on deep learning. Data Acquis. Process. 2022,
037.
156. Ouyu, C.; Yijun, L.; Wujian, Y.; Zhiwei, M.; Qi, L. Stroke prediction method based on deep learning and MFCC characteristics. In
Information and Computer (Theoretical Edition); 2019.
157. Kumar, S.; Sengupta, A. EEG Classification For Stroke Detection Using Deep Learning Networks. In Proceedings of the 2022 2nd
International Conference on Emerging Frontiers in Electrical and Electronic Technologies (ICEFEET), Patna, India, 24–25 June
2022; pp. 1–6.
158. Seal, A.; Bajpai, R.; Agnihotri, J.; Yazidi, A.; Herrera-Viedma, E.; Krejcar, O. DeprNet: A Deep Convolution Neural Network
Framework for Detecting Depression Using EEG. IEEE Trans. Instrum. Meas. 2021, 70, 1–13. [CrossRef]
159. Rafiei, A.; Zahedifar, R.; Sitaula, C.; Marzbanrad, F. Automated Detection of Major Depressive Disorder With EEG Signals: A
Time Series Classification Using Deep Learning. IEEE Access 2022, 10, 73804–73817. [CrossRef]
160. Sudhakar, T.; Hari Krishnan, G.; Krishnamoorthy, N.R.; Janney J, B.; Pradeepa, M.; Raghavi, J.P. Sleep Disorder Diagnosis using
EEG based Deep Learning Techniques. In Proceedings of the 2021 Seventh International conference on Bio Signals, Images, and
Instrumentation (ICBSII), Chennai, India, 25–27 March 2021; pp. 1–4.
161. Leino, A.; Korkalainen, H.; Kalevo, L.; Nikkonen, S.; Kainulainen, S.; Ryan, A.; Duce, B.; Sipilä, K.; Ahlberg, J.; Sahlman,
J.; et al. Deep Learning Enables Accurate Automatic Sleep Staging Based on Ambulatory Forehead EEG. IEEE Access 2022,
10, 26554–26566. [CrossRef]
Sensors 2023, 23, 6434 27 of 27

162. Kang, M.K.; Hong, K.S. A sleep stage classification method using deep learning by extracting the characteristics of frequency
domain from a single EEG channel. In Proceedings of the 2022 13th Asian Control Conference (ASCC), Jeju, Republic of Korea,
4–7 May 2022; pp. 1575–1580.
163. Almogbel, M.A.; Dang, A.H.; Kameyama, W. EEG-signals based cognitive workload detection of vehicle driver using deep
learning. In Proceedings of the 2018 20th International Conference on Advanced Communication Technology (ICACT),
Chuncheon, Republic of Korea, 11–14 February 2018; pp. 256–259.
164. Bhardwaj, R.; Parameswaran, S.; Balasubramanian, V. Performance Comparison of Machine Learning and Deep Learning While
Classifying Driver’s Cognitive State. In Proceedings of the 2018 IEEE 13th International Conference on Industrial and Information
Systems (ICIIS), Rupnagar, India, 1–2 December 2018; pp. 89–93.
165. Roy, A.D.; Islam, M.M. Detection of Epileptic Seizures from Wavelet Scalogram of EEG Signal Using Transfer Learning with
AlexNet Convolutional Neural Network. In Proceedings of the 2020 23rd International Conference on Computer and Information
Technology (ICCIT), Dhaka, Bangladesh, 19–21 December 2020; pp. 1–5.
166. Ak, A.; Topuz, V.; Midi, I. Motor imagery EEG signal classification using image processing technique over GoogLeNet deep
learning algorithm for controlling the robot manipulator. Biomed. Signal Process. Control 2022, 72, 103295. [CrossRef]
167. Bano, K.S.; Bhuyan, P.; Ray, A. EEG-Based Brain Computer Interface for Emotion Recognition. In Proceedings of the 2022 5th
International Conference on Computational Intelligence and Networks (CINE), Bhubaneswar, India, 1–3 December 2022; pp. 1–6.
168. Miao, Z.; Zhang, X.; Zhao, M.; Ming, D. LMDA-Net: A lightweight multi-dimensional attention network for general EEG-based
brain–computer interface paradigms and interpretability. arXiv 2023, arXiv:2303.16407.
169. Clerc, M.; Bougrain, L.; Lotte, F. Brain-Computer Interfaces 1: Methods and Perspectives; John Wiley & Sons: Hoboken, NJ, USA, 2016.
170. Corley, I.A.; Huang, Y. Deep EEG super-resolution: Upsampling EEG spatial resolution with generative adversarial networks. In
Proceedings of the 2018 IEEE EMBS International Conference on Biomedical & Health Informatics (BHI), Las Vegas, NV, USA, 4–7
March 2018; pp. 100–103.
171. Wang, F.; Zhong, S.H.; Peng, J.; Jiang, J.; Liu, Y. Data augmentation for EEG-based emotion recognition with deep convolutional
neural networks. In Proceedings of the MultiMedia Modeling: 24th International Conference, MMM 2018, Bangkok, Thailand,
5–7 February 2018; pp. 82–93.
172. Wan, Z.; Yang, R.; Huang, M.; Zeng, N.; Liu, X. A review on transfer learning in EEG signal analysis. Neurocomputing 2021,
421, 1–14. [CrossRef]
173. Wang, M.; Deng, W. Deep visual domain adaptation: A survey. Neurocomputing 2018, 312, 135–153. [CrossRef]
174. Wang, Y.; Qiu, S.; Li, D.; Du, C.; Lu, B.L.; He, H. Multi-modal domain adaptation variational autoencoder for eeg-based emotion
recognition. IEEE/CAA J. Autom. Sin. 2022, 9, 1612–1626. [CrossRef]
175. Voigt, P.; Von dem Bussche, A. The EU General Data Protection Regulation (GDPR); A Practical Guide; Springer: Cham, Switzerland,
2017; Volume 10, pp. 10–5555.
176. Chaddad, A.; Lu, Q.; Li, J.; Katib, Y.; Kateb, R.; Tanougast, C.; Bouridane, A.; Abdulkadir, A. Explainable, domain-adaptive, and
federated artificial intelligence in medicine. IEEE/CAA J. Autom. Sin. 2023, 10, 859–876. [CrossRef]
177. Sun, L.; Wu, J. A scalable and transferable federated learning system for classifying healthcare sensor data. IEEE J. Biomed. Health
Inform. 2022. [CrossRef] [PubMed]
178. Yildirim, O.; Baloglu, U.B.; Acharya, U.R. A deep learning model for automated sleep stages classification using PSG signals. Int.
J. Environ. Res. Public Health 2019, 16, 599. [CrossRef] [PubMed]

Disclaimer/Publisher’s Note: The statements, opinions and data contained in all publications are solely those of the individual
author(s) and contributor(s) and not of MDPI and/or the editor(s). MDPI and/or the editor(s) disclaim responsibility for any injury to
people or property resulting from any ideas, methods, instructions or products referred to in the content.