International Journal of Scientific Research & Engineering Trends

Volume 9, Issue 3, May-Jun-2023, ISSN (Online): 2395-566X

A Review of Motor Imagery EEG Classification Based on

Transform
M. Tech. Scholar Amitesh Raj, Prof. & HOD. Dr Bharti Chourasia
Department of Electronics & Communication Engineering,
RKDF Institute of Science & Technology SRK University
Bhopal(M.P)
Abstract- The biomedical play significant role in critical disease detection and prediction. The motor imagery-based EEG
classification is the approach to detection some serious disease related to human nervous system. The motor imagery EEG
signal data recoded with brain computer interface. The brain computer interface system equipped with electrode and sensors.
The recoded signals of nervous system are very complex and high dimension. The complexity of dimension and structure of
data face a problem of classification and detection. This paper presents the review of transform based methods of feature
extraction of motor imagery classification..

Keywords- EEG Classification, Machine learning, Motor Imagery.

I. INTRODUCTION exploiting Laplace priors, namely, SBLaplace, for EEG

classification. A sparse discriminant vector is learned with
Brain computer interface provides platform for the a Laplace prior in a hierarchical fashion under a Bayesian
analysis of human physical and kinetics behaviors evidence framework. All required model parameters are
analysis. Motor imagery based BCI is a very productive automatically estimated from training data without the
communication method for people with motor disabilities. need of CV.
Motor Imagery (MI) is a mental process wherein the
subject imagines that he is performing a specific motor Leonard J. Trejo, Karla Kubitz, Roman Rosipal,
action such as a hand or foot movement without otherwise Rebekah L. Kochavi and Leslie D. Montgomery Et al.
performing it in reality. Electroencephalogram (EEG) [2] Mental fatigue was associated with increased power in
signals are used as inputs to BCI systems. EEG signals frontal theta and parietal alpha EEG rhythms. A statistical
are feature extracted in order to overcome the classifier can use these effects to model EEG-fatigue
contaminations of noise and artifacts in them. relationships accurately. Mean power spectral densities or
PSDs and bands rose by 29% and 44%, respectively. A
Soft computing algorithms are then used in the kernel partial least squares classifier trained to classify
classification of different brain patterns obtained upon PSD coefficients (1 - 18 Hz) of single 13-s EEG segments
performing different motor imagery tasks. A BCI system from alert or fatigued task periods was 91% to 100%
measures brain activity and translates it into control accurate. For EEG segments from other task periods, the
signals. These control signals can be used to construct classifier outputs tracked the time course of the
new augmentative technologies. People with motor development of mental fatigue. By this measure, most
disabilities need augmentative technologies corresponding subjects became substantially fatigued after 60 min of
to natural ways of communications. In the rest part of his task performance.
paper, section II- literature survey. Section III-discussed
the problem domain and finally conclude the conclusions. Haider Raza, Hubert Cecotti and Girijesh Prasad Et
al. [3] A major problem in a brain-computer interface
II. LITERATURE SURVEY (BCI) based on electroencephalogram (EEG) recordings
is the varying statistical properties of the signals during
inter- or intra-session transfers that often lead to
Yu Zhang, Guoxu Zhou, Jing Jin, Qibin Zhao, Xingyu
Wang and Andrzej Cichocki Et al. [1] Regularization deteriorated BCI performances. A filter bank CSP
has been one of the most popular approaches to prevent (FBCSP) algorithm typically uses all the features from all
overfitting in electroencephalogram (EEG) classification the bands to extract and select robust features. In this
of brain–computer interfaces (BCIs). The effective-ness paper, they evaluate the performance of four methods for
of regularization is often highly dependent on the frequency band selection applied to binary motor imagery
selection of regularization parameters that are typically classification: forward- addition (FA), backward-
determined by cross- validation (CV). These limitations elimination (BE), the intersection and the union of the FA
substantially deteriorate the system’s practicability and and BE. These methods automatically select and learn the
may cause a user to be reluctant to use BCIs. In this best discriminative sets of frequency bands, and their
paper, they introduce a sparse Bayesian method by corresponding CSP features.

© 2023 IJSRET
722
 International Journal of Scientific Research & Engineering Trends
Volume 9, Issue 3, May-Jun-2023, ISSN (Online): 2395-566X

Jeong-Hwan Lim, Jun-Hak Lee, Han-Jeong Hwang, complexity of the learning algorithm so as to alleviate the
Dong Hwan Kim, Chang-Hwan Im. Et al. [4] They curse of dimensionality and attaining computational
introduced a hybrid mental spelling system which efficiency to facilitate online applications, e.g., brain
prevents additional typing of BACKSPACE to correct computer interfaces (BCIs). To tackle these barriers, this
typos. In order to detect typos, simultaneously utilizes paper presents a novel algorithm, termed regularized
both EEG signals recorded from the occipital area and the spatio-temporal filtering and classification (RSTFC), for
horizontal eye-gaze direction information extracted from single-trial EEG classification. RSTFC consists of two
a low-cost webcam-based eye tracker. In their online modules. In the feature extraction module, an l2-
experiments conducted with 10 healthy participants, at regularized algorithm is developed for supervised spatio-
least 16.6 typos could be prevented, from the results, temporal filtering of the EEG signals. Unlike the existing
verifying that the discussed strategy could effectively supervised spatio-temporal filter optimization algorithms,
enhance the performance of the SSVEP-based mental the developed algorithm can simultaneously optimize
spelling system. spatial and high-order temporal filters in an eigenvalue
decomposition framework and thus be implemented
Laura Acqualagna, Sebastian Bosse, Anne K highly efficiently.
Porbadnigk, Gabriel Curio, Klaus-Robert Müller,
Thomas Wiegand and Benjamin Blankertz Et al. [5] Minho Kim, Byung Hyung Kim and Sungho Jo Et al.
Recent studies exploit the neural signal recorded via [9] This paper describes a low-cost noninvasive brain-
electroencephalography (EEG) to get a more objective computer interface (BCI) hybridized with eye tracking. It
measurement of perceived video quality. Most of these also discusses its feasibility through a Fitts' law-based
studies capitalize on the event-related potential quantitative evaluation method. Noninvasive BCI has
component P3. They follow an alternative approach to the recently received a lot of attention. To bring the BCI
measurement problem investigating steady state visual applications into real life, user-friendly and easily
evoked potentials (SSVEPs) as EEG correlates of quality portable devices need to be provided. In this work, as an
changes. Unlike the P3, SSVEPs are directly linked to the approach to realize a real- world BCI,
sensory processing of the stimuli and do not require long electroencephalograph (EEG)-based BCI combined with
experimental sessions to get a sufficient signal-to-noise eye tracking is investigated.
ratio.
Oana Diana Eva and Anca Mihaela Lazar Et al. [10]
Dilshad Begum, K. M. Ravikumar, James. Mathew Using the EEG Motor Movement/Imagery database there
and Sanjeev Kubakaddi Et al. [6] Recent is discussed an off-line analysis for a brain computer
electrophysiological studies support command- specific interface (BCI) paradigm. The purpose of the quantitative
changes in the electroencephalography (EEG) that have research is to compare classifier in order to determinate
promoted their intensive application in the noninvasive which of them has highest rates of classification. The
brain computer interfaces (BCI). However, EEG is power spectral density method is used to evaluate the
plagued by a variety of interferences and noises, thereby (de)synchronizations that appear on Mu rhythm. The
demanding better accuracy and stability for its application features extracted from EEG signals are classified using
in the neuro-prosthetic devices. Here They investigate linear discriminant classifier (LDA), quadratic classifier
wavelets and adaptive neuro-fuzzy classification (QDA) and classifier based on Mahalanobis Distance
algorithms to enhance the classification accuracy of (MD). The differences between LDA, QDA and MD are
cognitive tasks. small, but the superiority of QDA was sustained by
analysis of variance (ANOVA).
Using a standard cognitive EEG dataset, James J. S.
Norton, Dong Sup Leeb, Jung Woo Leed, Woosik Lee, Chi Zhang, Li Tong, Ying Zeng, Jingfang Jiang,
Ohjin Kwon and Phillip Won Et al. [7] Recent Haibing Bu, Bin Yan and Jianxin Li Et al. [11]
advances in electrodes for noninvasive recording of Electroencephalogram (EEG) is susceptible to various
electroencephalograms expand opportunities collecting nonneural physiological artifacts. Automatic artifact
such data for diagnosis of neurological disorders and removal from EEG data remains a key challenge for
brain computer interfaces. Existing technologies, extracting relevant information from brain activities. To
however, cannot be used effectively in continuous, adapt to variable subjects and EEG acquisition
uninterrupted modes for more than a few days due to environments, this paper presents an automatic online
irritation and irreversible degradation in the electrical and artifact removal method based on a priori artifact
mechanical properties of the skin interface. information. The combination of discrete wavelet
transforms and independent component analysis (ICA),
Feifei Qi, Yuanqing Li and Wei Wu Et al. [8] Learning wavelet-ICA, was utilized to separate artifact
optimal spatio-temporal filters is a key to feature components. The artifact components were then
extraction for single-trial electroencephalogram (EEG) automatically identified using a priori artifact
classification. The challenges are controlling the information, which was acquired in advance.

© 2023 IJSRET
723
 International Journal of Scientific Research & Engineering Trends
Volume 9, Issue 3, May-Jun-2023, ISSN (Online): 2395-566X

Subsequently, signal reconstruction without artifact III. PROBLEM DOMAIN

components was performed to obtain artifact-free signals.
The extraction of features and band separation from the
Younghak Shin, Seungchan Lee, Minkyu Ahn, EEG signals are very difficult. The extraction of features
Hohyun Cho, Sung Chan Jun and Heung-No Lee Et plays important role in detection of some critical diseases
al. [12] In the electroencephalogram (EEG)-based brain- related to human brain
computer interface (BCI) systems, classification is an  Loss of features
important signal processing step to control external  Reduction of Nosie and interference
devices using brain activity. However, scalp-recorded  Separation of bands
EEG signals have inherent non-stationary characteristics;
thus, the classification performance is deteriorated by
changing the background activity of the EEG during the
IV. CONCLUSION
BCI experiment.
The motor imagery EEG signal classification is path of
Using the noisy test signals and real online-experimental complex diseases analysis in medical science. Now a
dataset, they compare the classification performance of day’s various authors and researcher used transform based
the SRC and support vector machine (SVM). technique for the categorization of EEG signals. The
Furthermore, they analyze the unique classification feature extraction and selection of features in EEG signal
mechanism of the SRC. They observed that the SRC play an important role. The extraction of features depends
method provided better classification accuracy and noise the raw data’s behavior.
robustness compared with the SVM method. In addition,
the SRC has an inherent adaptive classification The extraction of features decreases the raw information’s
mechanism that makes it suitable for time-varying EEG size of EEG signal. The wavelet transform based feature
signal classification for online BCI systems. extraction methods are better than other feature extraction
methods such as CSP and HFE. In future used hybrid
Huijuan Yang, Siavash Sakhavi, Kai Keng Ang and feature extraction algorithms for the removal of noise and
Cuntai Guan Et al. [13] they evaluated and analyzed the improve the performance of EEG signal classification.
robustness of the SRC method against the non-
stationarity of EEG signal classification. For this purpose, REFERENCES
they generated noise corrupted EEG test signals using two
noise sources such as random Gaussian noise and scalp [1] Yu Zhang, Guoxu Zhou, Jing Jin, Qibin Zhao,
recorded background noise. Then, they assessed the Xingyu Wang and Andrzej Cichocki “Sparse
classification performance of the SRC when the noise Bayesian Classification of EEG for Brain Computer
power was varied. Using the experimental motor imagery- Interface”, IEEE, 2020, Pp 1-13.
based EEG and generated noisy test signals, they [2] Leonard J. Trejo, Karla Kubitz, Roman Rosipal,
compared the classification results of the SRC with that of Rebekah L. Kochavi and Leslie D. Montgomery
the SVM method, which has been considered as a robust “EEG-Based Estimation and Classification of Mental
classifier in many BCI studies. Fatigue”, Psychology, 2019, Pp 572-589.
[3] Haider Raza, Hubert Cecotti and Girijesh Prasad
Ye Liu, Qibin Zhao and Liqing Zhang Et al. [14] “Optimising Frequency Band Selection with
Motor imagery-based brain computer interfaces (BCIs) Forward- Addition and Backward-Elimination
training has been proved to be an effective Algorithms in EEG-based Brain-Computer
communication system between human brain and external Interfaces”, IJCNN, 2019, Pp 1-8.
devices. A practical problem in BCI-based systems is how [4] Jeong-Hwan Lim, Jun-Hak Lee, Han-Jeong Hwang,
to correctly and efficiently identify and extract subject- Dong Hwan Kim, Chang-Hwan Im “Development of
specific features from the blurred scalp a hybrid mental spelling system combining SSVEP-
electroencephalography (EEG) and translate those based brain computer interface and webcam-based
features into device commands in order to control external eye tracking” Biomedical Signal Processing and
devices. In real BCI-based applications, they usually Control, 2019, Pp 99-104.
define frequency bands and channels configuration that [5] Laura Acqualagna, Sebastian Bosse, Anne K
related to brain activities beforehand. Porbadnigk, Gabriel Curio, Klaus-Robert Müller,
Thomas Wiegand and Benjamin Blankertz “EEG-
In this study, a robust tensor-based method is discussed based classification of video quality perception using
for a multiway discriminative subspace extraction from steady state visual evoked potentials (SSVEPs)”, J.
tensor-represented EEG data, which performs well in Neural Eng., 2020, Pp 1-17.
motor imagery EEG classification without the prior [6] Dilshad Begum, K. M. Ravikumar, James. Mathew
neurophysiologic knowledge like channels configuration and Sanjeev Kubakaddi “EEG Based Patient
and active frequency bands. Monitoring System for Mental Alertness Using

© 2023 IJSRET
724
 International Journal of Scientific Research & Engineering Trends
Volume 9, Issue 3, May-Jun-2023, ISSN (Online): 2395-566X

Adaptive Neuro-Fuzzy Approach”, Journal of

Medical and Bioengineering, 2019, Pp 59-66.
[7] James J. S. Norton, Dong Sup Leeb, Jung Woo Leed,
Woosik Lee, Ohjin Kwon and Phillip Won “Soft,
curved electrode systems capable of integration on
the auricle as a persistent brain computer interface”,
PNAS Early Edition, 2020, Pp 1-6.
[8] Feifei Qi, Yuanqing Li and Wei Wu “RSTFC: A
Novel Algorithm for Spatio-Temporal Filtering and
Classification of Single-Trial EEG”, IEEE, 2019, Pp
3070-3082.
[9] Minho Kim, Byung Hyung Kim and Sungho Jo
“Quantitative Evaluation of a Low-Cost Noninvasive
Hybrid Interface Based on EEG and Eye Movement”,
IEEE, 2018, Pp 159-168.
[10] Oana Diana Eva and Anca Mihaela Lazar
“Comparison of Classifiers and Statistical Analysis
for EEG Signals Used in Brain Computer Interface
Motor orporation Task Paradigm”, International
Journal of Advanced Research in Artificial
Intelligence, 2019, Pp 8-12.
[11] Chi Zhang, Li Tong, Ying Zeng, Jingfang Jiang,
Haibing Bu, Bin Yan and Jianxin Li “Automatic
Artifact Removal from Electroencephalogram Data
Based on A Priori Artifact Information”, Hindawi
Publishing Corporation, 2018, Pp 1-9.
[12] Younghak Shin, Seungchan Lee, Minkyu Ahn,
Hohyun Cho, Sung Chan Jun and Heung-No Lee
“Noise Robustness Analysis of Sparse Representation
based Classification Method for Non-stationary EEG
Signal Classification”, IEEE, 2020, Pp 1-10.
[13] Huijuan Yang, Siavash Sakhavi, Kai Keng Ang and
Cuntai Guan “On the Use of Convolutional Neural
Networks and Augmented CSP Features for Multi-
class Motor Imagery of EEG Signals Classification”,
IEEE, 2019, Pp 1-4.
[14] Ye Liu, Qibin Zhao and Liqing Zhang “Uncorrelated
Multiway Discriminant Analysis for Motor Imagery
EEG Classification”, International Journal of Neural
Systems, 2019, Pp 1-14.
[15] Amirhossein S. Aghaei, Mohammad Shahin Mahanta
and Konstantinos N. Plataniotis “Separable Common
Spatio-Spectral Patterns for Motor Imagery BCI
Systems”, IEEE, 2018, Pp 15-29.

© 2023 IJSRET
725