In this paper, different machine learning algorithms such as Linear Discriminant Analysis, Support vector
machine (SVM), Multi-layer perceptron, Random forest, K-nearest neighbour, and Autoencoder with SVM
have been compared. This comparison was conducted to seek a robust method that would produce good
classification accuracy. To this end, a robust method of classifying raw Electroencephalography (EEG)
signals associated with imagined movement of the right hand and relaxation state, namely Autoencoder
with SVM has been proposed. The EEG dataset used in this research was created by the University of
Tubingen, Germany. The best classification accuracy achieved was 70.4% with SVM through feature
engineering. However, our prosed method of autoencoder in combination with SVM produced a similar
accuracy of 65% without using any feature engineering technique. This research shows that this system of
classification of motor movements can be used in a Brain-Computer Interface system (BCI) to mentally
control a robotic device or an exoskeleton.
he main idea of the current work is to use a wireless Electroencephalography (EEG) headset as a remote control for the mouse cursor of a personal computer. The proposed system uses EEG signals as a communication link between brains and computers. Signal records obtained from the PhysioNet EEG dataset were analyzed using the Coif lets wavelets and many features were extracted using different amplitude estimators for the wavelet coefficients. The extracted features were inputted into machine learning algorithms to generate the decision rules required for our application. The suggested real time implementation of the system was tested and very good performance was achieved. This system could be helpful for disabled people as they can control computer applications via the imagination of fists and feet movements in addition to closing eyes for a short period of time
This document summarizes an approach for automated EEG artifact elimination using machine learning algorithms applied to features extracted from independent component analysis (ICA). The method involves preprocessing the EEG data with filtering and ICA, then generating features from the ICA results like topographic maps and power spectra. Machine learning classifiers like support vector machines and artificial neural networks are trained on the features to classify components as artifacts or EEG. The best performance was 95% accuracy using a neural network classifier trained on features from topographic map filtering and power spectra. This automated approach avoids time-consuming manual classification and enables real-time artifact removal.
CLASSIFICATION OF ELECTROENCEPHALOGRAM SIGNALS USING XGBOOST ALGORITHM AND SU...IRJET Journal
This document presents a study that uses the XGBoost algorithm and support vector machine to classify electroencephalogram (EEG) signals. The study acquires EEG data from healthy subjects and subjects with epilepsy during seizure and non-seizure periods. It preprocesses the data, extracts features using linear discriminant analysis, and feeds the extracted features into XGBoost and SVM classifiers. The results indicate that XGBoost exhibited superior classification performance compared to SVM for analyzing and classifying EEG signals.
Efficient electro encephelogram classification system using support vector ma...nooriasukmaningtyas
Complex modern signal processing is used to automate the analysis of electro encephelogram (EEG) signals. For the diagnosis of seizures, approaches that are simple and precise may be preferable rather than difficult and time-consuming. In this paper, efficient EEG classification system using support vector machine (SVM) and Adaptive learning technique is proposed. The database EEG signals are subjected to temporal and spatial filtering to remove unwanted noise and to increase the detection accuracy of the classifier by selecting the specific bands in which most of the EEG data are present. The neural network based SVM is used to classify the test EEG data with respect to training data. The cost-sensitive SVM with proposed Adaptive learning classifies the EEG signals where the adaptive learning with probability based function helps in prediction of the future samples and this leads in improving the accuracy with detection time. The detection accuracy of the proposed algorithm is compared with existing which shows that the proposed algorithm can classify the EEG signal more effectively.
Transfer learning for epilepsy detection using spectrogram imagesIAESIJAI
Epilepsy stands out as one of the common neurological diseases. The neural activity of the brain is observed using electroencephalography (EEG). Manual inspection of EEG brain signals is a slow and arduous process, which puts heavy load on neurologists and affects their performance. The aim of this study is to find the best result of classification using the transfer learning model that automatically identify the epileptic and the normal activity, to classify EEG signals by using images of spectrogram which represents the percentage of energy for each coefficient of the continuous wavelet. Dataset includes the EEG signals recorded at monitoring unit of epilepsy used in this study to presents an application of transfer learning by comparing three models Alexnet, visual geometry group (VGG19) and residual neural network ResNet using different combinations with seven different classifiers. This study tested the models and reached a different value of accuracy and other metrics used to judge their performances, and as a result the best combination has been achieved with ResNet combined with support vector machine (SVM) classifier that classified EEG signals with a high success rate using multiple performance metrics such as 97.22% accuracy and 2.78% the value of the error rate.
Technology Development for Unblessed people using BCI: A SurveyMandeep Kaur
This document summarizes research on brain-computer interface (BCI) systems that assist people with disabilities. It first reviews various BCI systems developed between 2000-2009 that use electroencephalography (EEG) signals to help people with conditions like paralysis. It then discusses applications of BCI for people with disabilities, including using mental commands to control assistive devices. Finally, it outlines the general components of a BCI system, including EEG signal acquisition, processing, feature extraction, classification, and applications. The overall purpose is to survey progress in developing BCIs to improve the lives of people with disabilities.
Improved feature exctraction process to detect seizure using CHBMIT-dataset ...IJECEIAES
One of the most dangerous neurological disease, which is occupying worldwide, is epilepsy. Fraction of second nerves in the brain starts impulsion i.e. electrical discharge, which is higher than the normal pulsing. So many researches have done the investigation and proposed the numerous methodology. However, our methodology will give effective result in feature extraction. Moreover, we used numerous number of statistical moments features. Existing approaches are implemented on few statistical moments with respect to time and frequency. Our proposed system will give the way to find out the seizure-effected part of the brain very easily using TDS, FDS, Correlation and Graph presentation. The resultant value will give the huge difference between normal and seizure effected brain. It also explore the hidden features of the brain.
Robot Motion Control Using the Emotiv EPOC EEG SystemjournalBEEI
Brain-computer interfaces have been explored for years with the intent of using human thoughts to control mechanical system. By capturing the transmission of signals directly from the human brain or electroencephalogram (EEG), human thoughts can be made as motion commands to the robot. This paper presents a prototype for an electroencephalogram (EEG) based brain-actuated robot control system using mental commands. In this study, Linear Discriminant Analysis (LDA) and Support Vector Machine (SVM) method were combined to establish the best model. Dataset containing features of EEG signals were obtained from the subject non-invasively using Emotiv EPOC headset. The best model was then used by Brain-Computer Interface (BCI) to classify the EEG signals into robot motion commands to control the robot directly. The result of the classification gave the average accuracy of 69.06%.
he main idea of the current work is to use a wireless Electroencephalography (EEG) headset as a remote control for the mouse cursor of a personal computer. The proposed system uses EEG signals as a communication link between brains and computers. Signal records obtained from the PhysioNet EEG dataset were analyzed using the Coif lets wavelets and many features were extracted using different amplitude estimators for the wavelet coefficients. The extracted features were inputted into machine learning algorithms to generate the decision rules required for our application. The suggested real time implementation of the system was tested and very good performance was achieved. This system could be helpful for disabled people as they can control computer applications via the imagination of fists and feet movements in addition to closing eyes for a short period of time
This document summarizes an approach for automated EEG artifact elimination using machine learning algorithms applied to features extracted from independent component analysis (ICA). The method involves preprocessing the EEG data with filtering and ICA, then generating features from the ICA results like topographic maps and power spectra. Machine learning classifiers like support vector machines and artificial neural networks are trained on the features to classify components as artifacts or EEG. The best performance was 95% accuracy using a neural network classifier trained on features from topographic map filtering and power spectra. This automated approach avoids time-consuming manual classification and enables real-time artifact removal.
CLASSIFICATION OF ELECTROENCEPHALOGRAM SIGNALS USING XGBOOST ALGORITHM AND SU...IRJET Journal
This document presents a study that uses the XGBoost algorithm and support vector machine to classify electroencephalogram (EEG) signals. The study acquires EEG data from healthy subjects and subjects with epilepsy during seizure and non-seizure periods. It preprocesses the data, extracts features using linear discriminant analysis, and feeds the extracted features into XGBoost and SVM classifiers. The results indicate that XGBoost exhibited superior classification performance compared to SVM for analyzing and classifying EEG signals.
Efficient electro encephelogram classification system using support vector ma...nooriasukmaningtyas
Complex modern signal processing is used to automate the analysis of electro encephelogram (EEG) signals. For the diagnosis of seizures, approaches that are simple and precise may be preferable rather than difficult and time-consuming. In this paper, efficient EEG classification system using support vector machine (SVM) and Adaptive learning technique is proposed. The database EEG signals are subjected to temporal and spatial filtering to remove unwanted noise and to increase the detection accuracy of the classifier by selecting the specific bands in which most of the EEG data are present. The neural network based SVM is used to classify the test EEG data with respect to training data. The cost-sensitive SVM with proposed Adaptive learning classifies the EEG signals where the adaptive learning with probability based function helps in prediction of the future samples and this leads in improving the accuracy with detection time. The detection accuracy of the proposed algorithm is compared with existing which shows that the proposed algorithm can classify the EEG signal more effectively.
Transfer learning for epilepsy detection using spectrogram imagesIAESIJAI
Epilepsy stands out as one of the common neurological diseases. The neural activity of the brain is observed using electroencephalography (EEG). Manual inspection of EEG brain signals is a slow and arduous process, which puts heavy load on neurologists and affects their performance. The aim of this study is to find the best result of classification using the transfer learning model that automatically identify the epileptic and the normal activity, to classify EEG signals by using images of spectrogram which represents the percentage of energy for each coefficient of the continuous wavelet. Dataset includes the EEG signals recorded at monitoring unit of epilepsy used in this study to presents an application of transfer learning by comparing three models Alexnet, visual geometry group (VGG19) and residual neural network ResNet using different combinations with seven different classifiers. This study tested the models and reached a different value of accuracy and other metrics used to judge their performances, and as a result the best combination has been achieved with ResNet combined with support vector machine (SVM) classifier that classified EEG signals with a high success rate using multiple performance metrics such as 97.22% accuracy and 2.78% the value of the error rate.
Technology Development for Unblessed people using BCI: A SurveyMandeep Kaur
This document summarizes research on brain-computer interface (BCI) systems that assist people with disabilities. It first reviews various BCI systems developed between 2000-2009 that use electroencephalography (EEG) signals to help people with conditions like paralysis. It then discusses applications of BCI for people with disabilities, including using mental commands to control assistive devices. Finally, it outlines the general components of a BCI system, including EEG signal acquisition, processing, feature extraction, classification, and applications. The overall purpose is to survey progress in developing BCIs to improve the lives of people with disabilities.
Improved feature exctraction process to detect seizure using CHBMIT-dataset ...IJECEIAES
One of the most dangerous neurological disease, which is occupying worldwide, is epilepsy. Fraction of second nerves in the brain starts impulsion i.e. electrical discharge, which is higher than the normal pulsing. So many researches have done the investigation and proposed the numerous methodology. However, our methodology will give effective result in feature extraction. Moreover, we used numerous number of statistical moments features. Existing approaches are implemented on few statistical moments with respect to time and frequency. Our proposed system will give the way to find out the seizure-effected part of the brain very easily using TDS, FDS, Correlation and Graph presentation. The resultant value will give the huge difference between normal and seizure effected brain. It also explore the hidden features of the brain.
Robot Motion Control Using the Emotiv EPOC EEG SystemjournalBEEI
Brain-computer interfaces have been explored for years with the intent of using human thoughts to control mechanical system. By capturing the transmission of signals directly from the human brain or electroencephalogram (EEG), human thoughts can be made as motion commands to the robot. This paper presents a prototype for an electroencephalogram (EEG) based brain-actuated robot control system using mental commands. In this study, Linear Discriminant Analysis (LDA) and Support Vector Machine (SVM) method were combined to establish the best model. Dataset containing features of EEG signals were obtained from the subject non-invasively using Emotiv EPOC headset. The best model was then used by Brain-Computer Interface (BCI) to classify the EEG signals into robot motion commands to control the robot directly. The result of the classification gave the average accuracy of 69.06%.
Electroencephalography-based brain-computer interface using neural networksTELKOMNIKA JOURNAL
This study aimed to develop a brain-computer interface that can control an electric wheelchair using electroencephalography (EEG) signals. First, we used the Mind Wave Mobile 2 device to capture raw EEG signals from the surface of the scalp. The signals were transformed into the frequency domain using fast Fourier transform (FFT) and filtered to monitor changes in attention and relaxation. Next, we performed time and frequency domain analyses to identify features for five eye gestures: opened, closed, blink per second, double blink, and lookup. The base state was the opened-eyes gesture, and we compared the features of the remaining four action gestures to the base state to identify potential gestures. We then built a multilayer neural network to classify these features into five signals that control the wheelchair’s movement. Finally, we designed an experimental wheelchair system to test the effectiveness of the proposed approach. The results demonstrate that the EEG classification was highly accurate and computationally efficient. Moreover, the average performance of the brain-controlled wheelchair system was over 75% across different individuals, which suggests the feasibility of this approach.
Survey analysis for optimization algorithms applied to electroencephalogramIJECEIAES
This paper presents a survey for optimization approaches that analyze and classify electroencephalogram (EEG) signals. The automatic analysis of EEG presents a significant challenge due to the high-dimensional data volume. Optimization algorithms seek to achieve better accuracy by selecting practical features and reducing unwanted features. Forty-seven reputable research papers are provided in this work, emphasizing the developed and executed techniques divided into seven groups based on the applied optimization algorithm particle swarm optimization (PSO), ant colony optimization (ACO), artificial bee colony (ABC), grey wolf optimizer (GWO), Bat, Firefly, and other optimizer approaches). The main measures to analyze this paper are accuracy, precision, recall, and F1-score assessment. Several datasets have been utilized in the included papers like EEG Bonn University, CHB-MIT, electrocardiography (ECG) dataset, and other datasets. The results have proven that the PSO and GWO algorithms have achieved the highest accuracy rate of around 99% compared with other techniques.
Motor Imagery based Brain Computer Interface for Windows Operating SystemIRJET Journal
This document summarizes a research paper that proposes a motor imagery-based brain computer interface (MI-BCI) to allow physically challenged individuals to control basic functions of a Windows operating system using only their brain activity. The MI-BCI system uses an 8-channel EEG device to capture brainwaves while a subject imagines moving their arm or blinking their eyes. A convolutional neural network (CNN) classifier is trained on the EEG data to identify 7 possible commands: left, right, up, down mouse movement or left/right click, or an idle state. The trained CNN achieved an average accuracy of 92.85% in identifying commands. A Python program integrates the EEG data stream, CNN classifier, and Windows mouse/
Effective electroencephalogram based epileptic seizure detection using suppo...IJECEIAES
Epilepsy is one of the widespread disorders. It is a noncommunicable disease that affects the human nerve system. Seizures are abnormal patterns of behavior in the electricity of the brain which produce symptoms like losing consciousness, attention or convulsions in the whole body. This paper demonstrates an effective electroencephalogram (EEG) based seizure detection method using discrete wavelet transformation (DWT) for signal decomposition to extract features. An automatic channel selection method was proposed by the researcher to select the best channel from 23 channels based on maximum variance value. The records were segmented into a nonoverlapping segment with long 1-S. The support vector machine (SVM) model was used to automatically detect segments that contain seizures, using both frequency and time domain statistical moment features. The experimental result was obtained from 24 patients in CHB-MIT database. The average accuracy is 94.1, sensitivity is 93.5, specificity is 94.6 and the false positive rate average is 0.054.
Impact of adaptive filtering-based component analysis method on steady-state ...IAESIJAI
The significance of brain computer interface (BCI) systems is immensely high, especially for disabled people and patients with nervous system failure. Therefore, in this study, adaptive filtering-based component analysis (AFCA) model is presented to enhance target box identification efficiency at varied flickering frequencies in a visual stimulation process by efficient acquisition of electroencephalogram (EEG) signals for the application of steady-state visually evoked potential based BCI system. Furthermore, optimization of proposed AFCA model is performed based on the maximized reproducibility of correlated components. A multimedia authoring and management using your eyes and mind (MAMEM) steady-state visual evoked potential (SSVEP) dataset is utilized for efficient training of EEG signals and background entities are eliminated using adaptive filters in a pre-processing stage. Additionally, spatial filtering components are obtained to detect target flickering box based on the obtained quality features. Performance is measured by acquisition of SSVEP signals in terms of reconstruction efficiency, classification accuracy and information transfer rate (ITR) using proposed AFCA model. Mean classification accuracy for all 11 subject is 93.48% and ITR is 308.23 bpm. Further, classification accuracy is relatively higher than various SSVEP classification algorithms.
Study and analysis of motion artifacts for ambulatory electroencephalographyIJECEIAES
Motion artifacts contribute complexity in acquiring clean electroencephalography (EEG) data. It is one of the major challenges for ambulatory EEG. The performance of mobile health monitoring, neurological disorders diagnosis and surgeries can be significantly improved by reducing the motion artifacts. Although different papers have proposed various novel approaches for removing motion artifacts, the datasets used to validate those algorithms are questionable. In this paper, a unique EEG dataset was presented where ten different activities were performed. No such previous EEG recordings using EMOTIV EEG headset are available in research history that explicitly mentioned and considered a number of daily activities that induced motion artifacts in EEG recordings. Quantitative study shows that in comparison to correlation coefficient, the coherence analysis depicted a better similarity measure between motion artifacts and motion sensor data. Motion artifacts were characterized with very low frequency which overlapped with the Delta rhythm of the EEG. Also, a general wavelet transform based approach was presented to remove motion artifacts. Further experiment and analysis with more similarity metrics and longer recording duration for each activity is required to finalize the characteristics of motion artifacts and henceforth reliably identify and subsequently remove the motion artifacts in the contaminated EEG recordings.
Health electroencephalogram epileptic classification based on Hilbert probabi...IJECEIAES
This paper has proposed a new classification method based on Hilbert probability similarity to detect epileptic seizures from electroencephalogram (EEG) signals. Hilbert similarity probability-based measure is exploited to measure the similarity between signals. The proposed system consisted of models based on Hilbert probability similarity (HPS) to predict the state for the specific EEG signal. Particle swarm optimization (PSO) has been employed for feature selection and extraction. Furthermore, the used dataset in this study is Bonn University's publicly available EEG dataset. Several metrics are calculated to assess the performance of the suggested systems such as accuracy, precision, recall, and F1-score. The experimental results show that the suggested model is an effective tool for classifying EEG signals, with an accuracy of up to 100% for two-class status.
Brain computer interfacing for controlling wheelchair movementIRJET Journal
This document describes research on developing a brain-computer interface (BCI) system to control a wheelchair using EEG brain wave signals. Specifically, it focuses on using alpha waves detected during a relaxed state to allow users to control wheelchair movement and direction. The system is intended to help people with disabilities who cannot move themselves. The document provides background on BCI and previous related work, then describes the proposed system which uses EEG signals from a low-cost headset to classify motor imagery and control a wheelchair wirelessly. It discusses the algorithms and experimental results, showing the system can accurately detect different movement intentions based on alpha wave detection with minimal training.
Classification of electroencephalography using cooperative learning based on...IJECEIAES
Modern technologies are widely used today to diagnose epilepsy, neurological disorders, and brain tumors. Meanwhile, it is not cost-effective in terms of time and money to use a large amount of electroencephalography (EEG) data from different centers and collect them in a central server for processing and analysis. Collecting this data correctly is challenging, and organizations avoid sharing their and client information with others due to data privacy protection. It is difficult to collect these data correctly and it is challenging to transfer them to research centers due to the privacy of the data. In this regard, collaborative learning as an extraordinary approach in this field paves the way for the use of information repositories in research matters without transferring the original data to the centers. This study focuses on the use of a heterogeneous client balancing technique with an interval selection approach and classification of EEG signals with ResNet50 deep architecture. The test results achieved an accuracy of 99.14 compared to similar methods.
Brain computer interface based smart keyboard using neurosky mindwave headsetTELKOMNIKA JOURNAL
This document describes a brain-computer interface (BCI) system that uses a Neurosky Mindwave headset to detect brain signals and control a virtual keyboard. The system collects EEG data in real-time from the headset, analyzes it to extract attention and blink features, and uses those features to scan and select characters on the virtual keyboard. An experiment tested the system on 5 users over multiple sessions and found encouraging results, with users achieving text entry speeds of 1.55-1.8 words per minute, faster than some other BCI keyboard studies.
Implementation and Evaluation of Signal Processing Techniques for EEG based B...Damian Quinn
This document compares two approaches for classifying EEG signals from a brain-computer interface - a multi-layer perceptron neural network with Levenberg-Marquardt learning, and an Adaptive Neuro-Fuzzy Inference System. It analyzes EEG data from a dataset involving motor imagery tasks of left and right hand movement. Features are extracted from the EEG signals and both the neural network and ANFIS are used to classify the signals based on the features. The performance of the two classification approaches are then compared to determine if the hybrid ANFIS method can outperform the established neural network approach.
Classification of emotions induced by horror and relaxing movies using single-...IJECEIAES
It has been observed from recent studies that corticolimbic Theta rhythm from EEG recordings perceived as fear or threatening scene during neural processing of visual stimuli. In additions, neural oscillations’ patterns in Theta, Alpha and Beta sub-bands also play important role in brain’s emotional processing. Inspired from these findings, in this paper we attempt to classify two different emotional states by analyzing single-channel EEG recordings. A video clip that can evoke 3 different emotional states: neutral, relaxation and scary is shown to 19 college-aged subjects and they were asked to score their emotional outcome by giving a number between 0 to 10 (where 0 means not scary at all and 10 means the most scary). First, recorded EEG data were preprocessed by stationary wavelet transform (SWT) based artifact removal algorithm. Then power distribution in simultaneous time-frequency domain was analyzed using short-time Fourier transform (STFT) followed by calculating the average power during each 0.2s time-segment for each brain sub-band. Finally, 46 features, as the mean power of frequency bands between 4 and 50 Hz during each time-segment, containing 689 instances—for each subject —were collected for classification. We found that relaxation and fear emotions evoked during watching scary and relaxing movies can be classified with average classification rate of 94.208% using K-NN by applying methods and materials proposed in this paper. We also classified the dataset using SVM and we found out that K-NN classifier (when k = 1) outperforms SVM in classifying EEG dynamics induced by horror and relaxing movies, however, for K > 1 in K-NN, SVM has better average classification rate.
Overview of Machine Learning and Deep Learning Methods in Brain Computer Inte...IJCSES Journal
Research under the field of Brain Computer Interfaces is adapting various Machine Learning and Deep
Learning techniques in recent times. With the advent of modern BCI, the data generated by various devices
is now capable of detecting brain signals more accurately. This paper gives an overview of all the steps
involved in the process of applying Machine Learning as well as Deep Learning methods from Data
Acquisition to application of algorithms. It aims to study techniques currently employed to extract data,
features from brain data, different algorithms employed to draw insights from the extracted features, and
how it can be used in various BCI applications. By this study, I aim to put forward current Machine
Learning and Deep Learning Trends in the field of BCI.
International Journal of Engineering Research and Applications (IJERA) is an open access online peer reviewed international journal that publishes research and review articles in the fields of Computer Science, Neural Networks, Electrical Engineering, Software Engineering, Information Technology, Mechanical Engineering, Chemical Engineering, Plastic Engineering, Food Technology, Textile Engineering, Nano Technology & science, Power Electronics, Electronics & Communication Engineering, Computational mathematics, Image processing, Civil Engineering, Structural Engineering, Environmental Engineering, VLSI Testing & Low Power VLSI Design etc.
EEG S IGNAL Q UANTIFICATION B ASED ON M ODUL L EVELS sipij
This article proposes a contribution to quantify EE
G signals outline. This technique uses two tools fo
r EEG
signal characteristics extraction. Our tests were r
ealized on the basis of 32 canals EEG canals using
Neuroscan software. EEG example demonstration is re
ferenced CZ and is sampled at 1000HZ. The
principal aim of this technique is to reduce the im
portant volume of EEG signal data Without losing an
y
information. EEG signals are quantified on the basi
s of a whole predefined levels The obtained results
show that an EEG alignment can be posted in a quant
ified form.
A LOW COST EEG BASED BCI PROSTHETIC USING MOTOR IMAGERY ijitcs
Brain Computer Interfaces (BCI) provide the opportunity to control external devices using the brain
ElectroEncephaloGram (EEG) signals. In this paper we propose two software framework in order to
control a 5 degree of freedom robotic and prosthetic hand. Results are presented where an Emotiv
Cognitive Suite (i.e. the 1st framework) combined with an embedded software system (i.e. an open source
Arduino board) is able to control the hand through character input associated with the taught actions of
the suite. This system provides evidence of the feasibility of brain signals being a viable approach to
controlling the chosen prosthetic. Results are then presented in the second framework. This latter one
allowed for the training and classification of EEG signals for motor imagery tasks. When analysing the
system, clear visual representations of the performance and accuracy are presented in the results using a
confusion matrix, accuracy measurement and a feedback bar signifying signal strength. Experiments with
various acquisition datasets were carried out and with a critical evaluation of the results given. Finally
depending on the classification of the brain signal a Python script outputs the driving command to the
Arduino to control the prosthetic. The proposed architecture performs overall good results for the design
and implementation of economically convenient BCI and prosthesis.
A machine learning algorithm for classification of mental tasks.pdfPravinKshirsagar11
In this article, a contemporary tack of mental tasks on cognitive parts of humans is appraised using two different approaches such as wavelet transforms at a discrete time (DWT) and support vector machine (SVM). The put forth tack is instilled with the electroencephalogram (EEG) database acquired in real-time from CARE Hospital, Nagpur. Additional data is also acquired from a brain-computer interface (BCI). In the working model, signals from the database are wed out into different frequency sub-bands using DWT. Initially, updated statistical features are obtained from different frequency sub-bands. This type of representation defines the wavelet co-efficient which is introduced for reducing the measurement of data. Then, the projected method is realized using SVM for segregating both port and veracious hand movement. After segregation of EEG signals, results are achieved with an accuracy of 92% for BCI competition paradigm III and 97.89% for B-alert machine.
The document describes a brain-computer interface (BCI) system that uses electroencephalography (EEG) to classify motor imagery of the left or right arm and control an assistive device for paralyzed upper limbs. EEG signals are recorded over motor cortex areas during right and left arm imagery tasks. The mu and beta frequency bands are extracted and used to classify intended movement based on features like power and mean. If right arm imagery is classified, a stepper motor attached to the patient's forearm is activated to help lift their arm. The system was tested on 7 subjects with over 10 trials each, achieving classification of intended movement.
EEG SIGNAL CLASSIFICATION USING LDA AND MLP CLASSIFIERhiij
Electroencephalography (EEG) is the recording of electrical activities along the scalp. EEG measures
voltage fluctuations resulting from ionic current flows within the neurons of the brain. Diagnostic
applications generally focus on the spectral content of EEG, which is the type of neural oscillations that
can be observed in EEG signal. EEG is most often used to diagnose epilepsy, which causes obvious
abnormalities in EEG readings. This powerful property confirms the rich potential for EEG analysis and
motivates the need for advanced signal processing techniques to aid clinicians in their interpretations.
This paper describes the application of Wavelet Transform (WT) for the processing of
Electroencephalogram (EEG) signals. Furthermore, the linear discriminant analysis (LDA) is applied for
feature selection and dimensionality reduction where the informative and discriminative two-dimension
features are used as a benchmark for classification purposes through a Multi-Layers Perceptron (MLP)
neural network. For five classification problems, the proposed model achieves a high sensitivity,
specificity and accuracy of 100%.Finally, the comparison of the results obtained with the proposed
methods and those obtained with previous literature methods shows the superiority of our approach for
EEG signals classification and automated diagnosis
Sample size determination for classification of eeg signals using power analy...iaemedu
The document discusses determining the minimum sample size needed for classification of electroencephalogram (EEG) signals using machine learning. It proposes using power analysis to calculate the required sample size to separate classes with statistical stability. Power analysis was performed on a dataset of 500 EEG signals from 5 classes. The results found that a sample size of 81 signals is needed to achieve 95% power. Additional experiments varied the power level and error probability to relate their effects on minimum sample size. The sample sizes calculated from power analysis were validated using a decision tree classifier on the EEG dataset.
We have compiled the most important slides from each speaker's presentation. This year’s compilation, available for free, captures the key insights and contributions shared during the DfMAy 2024 conference.
DEEP LEARNING FOR SMART GRID INTRUSION DETECTION: A HYBRID CNN-LSTM-BASED MODELgerogepatton
As digital technology becomes more deeply embedded in power systems, protecting the communication
networks of Smart Grids (SG) has emerged as a critical concern. Distributed Network Protocol 3 (DNP3)
represents a multi-tiered application layer protocol extensively utilized in Supervisory Control and Data
Acquisition (SCADA)-based smart grids to facilitate real-time data gathering and control functionalities.
Robust Intrusion Detection Systems (IDS) are necessary for early threat detection and mitigation because
of the interconnection of these networks, which makes them vulnerable to a variety of cyberattacks. To
solve this issue, this paper develops a hybrid Deep Learning (DL) model specifically designed for intrusion
detection in smart grids. The proposed approach is a combination of the Convolutional Neural Network
(CNN) and the Long-Short-Term Memory algorithms (LSTM). We employed a recent intrusion detection
dataset (DNP3), which focuses on unauthorized commands and Denial of Service (DoS) cyberattacks, to
train and test our model. The results of our experiments show that our CNN-LSTM method is much better
at finding smart grid intrusions than other deep learning algorithms used for classification. In addition,
our proposed approach improves accuracy, precision, recall, and F1 score, achieving a high detection
accuracy rate of 99.50%.
More Related Content
Similar to A Comparative Study of Machine Learning Algorithms for EEG Signal Classification
Electroencephalography-based brain-computer interface using neural networksTELKOMNIKA JOURNAL
This study aimed to develop a brain-computer interface that can control an electric wheelchair using electroencephalography (EEG) signals. First, we used the Mind Wave Mobile 2 device to capture raw EEG signals from the surface of the scalp. The signals were transformed into the frequency domain using fast Fourier transform (FFT) and filtered to monitor changes in attention and relaxation. Next, we performed time and frequency domain analyses to identify features for five eye gestures: opened, closed, blink per second, double blink, and lookup. The base state was the opened-eyes gesture, and we compared the features of the remaining four action gestures to the base state to identify potential gestures. We then built a multilayer neural network to classify these features into five signals that control the wheelchair’s movement. Finally, we designed an experimental wheelchair system to test the effectiveness of the proposed approach. The results demonstrate that the EEG classification was highly accurate and computationally efficient. Moreover, the average performance of the brain-controlled wheelchair system was over 75% across different individuals, which suggests the feasibility of this approach.
Survey analysis for optimization algorithms applied to electroencephalogramIJECEIAES
This paper presents a survey for optimization approaches that analyze and classify electroencephalogram (EEG) signals. The automatic analysis of EEG presents a significant challenge due to the high-dimensional data volume. Optimization algorithms seek to achieve better accuracy by selecting practical features and reducing unwanted features. Forty-seven reputable research papers are provided in this work, emphasizing the developed and executed techniques divided into seven groups based on the applied optimization algorithm particle swarm optimization (PSO), ant colony optimization (ACO), artificial bee colony (ABC), grey wolf optimizer (GWO), Bat, Firefly, and other optimizer approaches). The main measures to analyze this paper are accuracy, precision, recall, and F1-score assessment. Several datasets have been utilized in the included papers like EEG Bonn University, CHB-MIT, electrocardiography (ECG) dataset, and other datasets. The results have proven that the PSO and GWO algorithms have achieved the highest accuracy rate of around 99% compared with other techniques.
Motor Imagery based Brain Computer Interface for Windows Operating SystemIRJET Journal
This document summarizes a research paper that proposes a motor imagery-based brain computer interface (MI-BCI) to allow physically challenged individuals to control basic functions of a Windows operating system using only their brain activity. The MI-BCI system uses an 8-channel EEG device to capture brainwaves while a subject imagines moving their arm or blinking their eyes. A convolutional neural network (CNN) classifier is trained on the EEG data to identify 7 possible commands: left, right, up, down mouse movement or left/right click, or an idle state. The trained CNN achieved an average accuracy of 92.85% in identifying commands. A Python program integrates the EEG data stream, CNN classifier, and Windows mouse/
Effective electroencephalogram based epileptic seizure detection using suppo...IJECEIAES
Epilepsy is one of the widespread disorders. It is a noncommunicable disease that affects the human nerve system. Seizures are abnormal patterns of behavior in the electricity of the brain which produce symptoms like losing consciousness, attention or convulsions in the whole body. This paper demonstrates an effective electroencephalogram (EEG) based seizure detection method using discrete wavelet transformation (DWT) for signal decomposition to extract features. An automatic channel selection method was proposed by the researcher to select the best channel from 23 channels based on maximum variance value. The records were segmented into a nonoverlapping segment with long 1-S. The support vector machine (SVM) model was used to automatically detect segments that contain seizures, using both frequency and time domain statistical moment features. The experimental result was obtained from 24 patients in CHB-MIT database. The average accuracy is 94.1, sensitivity is 93.5, specificity is 94.6 and the false positive rate average is 0.054.
Impact of adaptive filtering-based component analysis method on steady-state ...IAESIJAI
The significance of brain computer interface (BCI) systems is immensely high, especially for disabled people and patients with nervous system failure. Therefore, in this study, adaptive filtering-based component analysis (AFCA) model is presented to enhance target box identification efficiency at varied flickering frequencies in a visual stimulation process by efficient acquisition of electroencephalogram (EEG) signals for the application of steady-state visually evoked potential based BCI system. Furthermore, optimization of proposed AFCA model is performed based on the maximized reproducibility of correlated components. A multimedia authoring and management using your eyes and mind (MAMEM) steady-state visual evoked potential (SSVEP) dataset is utilized for efficient training of EEG signals and background entities are eliminated using adaptive filters in a pre-processing stage. Additionally, spatial filtering components are obtained to detect target flickering box based on the obtained quality features. Performance is measured by acquisition of SSVEP signals in terms of reconstruction efficiency, classification accuracy and information transfer rate (ITR) using proposed AFCA model. Mean classification accuracy for all 11 subject is 93.48% and ITR is 308.23 bpm. Further, classification accuracy is relatively higher than various SSVEP classification algorithms.
Study and analysis of motion artifacts for ambulatory electroencephalographyIJECEIAES
Motion artifacts contribute complexity in acquiring clean electroencephalography (EEG) data. It is one of the major challenges for ambulatory EEG. The performance of mobile health monitoring, neurological disorders diagnosis and surgeries can be significantly improved by reducing the motion artifacts. Although different papers have proposed various novel approaches for removing motion artifacts, the datasets used to validate those algorithms are questionable. In this paper, a unique EEG dataset was presented where ten different activities were performed. No such previous EEG recordings using EMOTIV EEG headset are available in research history that explicitly mentioned and considered a number of daily activities that induced motion artifacts in EEG recordings. Quantitative study shows that in comparison to correlation coefficient, the coherence analysis depicted a better similarity measure between motion artifacts and motion sensor data. Motion artifacts were characterized with very low frequency which overlapped with the Delta rhythm of the EEG. Also, a general wavelet transform based approach was presented to remove motion artifacts. Further experiment and analysis with more similarity metrics and longer recording duration for each activity is required to finalize the characteristics of motion artifacts and henceforth reliably identify and subsequently remove the motion artifacts in the contaminated EEG recordings.
Health electroencephalogram epileptic classification based on Hilbert probabi...IJECEIAES
This paper has proposed a new classification method based on Hilbert probability similarity to detect epileptic seizures from electroencephalogram (EEG) signals. Hilbert similarity probability-based measure is exploited to measure the similarity between signals. The proposed system consisted of models based on Hilbert probability similarity (HPS) to predict the state for the specific EEG signal. Particle swarm optimization (PSO) has been employed for feature selection and extraction. Furthermore, the used dataset in this study is Bonn University's publicly available EEG dataset. Several metrics are calculated to assess the performance of the suggested systems such as accuracy, precision, recall, and F1-score. The experimental results show that the suggested model is an effective tool for classifying EEG signals, with an accuracy of up to 100% for two-class status.
Brain computer interfacing for controlling wheelchair movementIRJET Journal
This document describes research on developing a brain-computer interface (BCI) system to control a wheelchair using EEG brain wave signals. Specifically, it focuses on using alpha waves detected during a relaxed state to allow users to control wheelchair movement and direction. The system is intended to help people with disabilities who cannot move themselves. The document provides background on BCI and previous related work, then describes the proposed system which uses EEG signals from a low-cost headset to classify motor imagery and control a wheelchair wirelessly. It discusses the algorithms and experimental results, showing the system can accurately detect different movement intentions based on alpha wave detection with minimal training.
Classification of electroencephalography using cooperative learning based on...IJECEIAES
Modern technologies are widely used today to diagnose epilepsy, neurological disorders, and brain tumors. Meanwhile, it is not cost-effective in terms of time and money to use a large amount of electroencephalography (EEG) data from different centers and collect them in a central server for processing and analysis. Collecting this data correctly is challenging, and organizations avoid sharing their and client information with others due to data privacy protection. It is difficult to collect these data correctly and it is challenging to transfer them to research centers due to the privacy of the data. In this regard, collaborative learning as an extraordinary approach in this field paves the way for the use of information repositories in research matters without transferring the original data to the centers. This study focuses on the use of a heterogeneous client balancing technique with an interval selection approach and classification of EEG signals with ResNet50 deep architecture. The test results achieved an accuracy of 99.14 compared to similar methods.
Brain computer interface based smart keyboard using neurosky mindwave headsetTELKOMNIKA JOURNAL
This document describes a brain-computer interface (BCI) system that uses a Neurosky Mindwave headset to detect brain signals and control a virtual keyboard. The system collects EEG data in real-time from the headset, analyzes it to extract attention and blink features, and uses those features to scan and select characters on the virtual keyboard. An experiment tested the system on 5 users over multiple sessions and found encouraging results, with users achieving text entry speeds of 1.55-1.8 words per minute, faster than some other BCI keyboard studies.
Implementation and Evaluation of Signal Processing Techniques for EEG based B...Damian Quinn
This document compares two approaches for classifying EEG signals from a brain-computer interface - a multi-layer perceptron neural network with Levenberg-Marquardt learning, and an Adaptive Neuro-Fuzzy Inference System. It analyzes EEG data from a dataset involving motor imagery tasks of left and right hand movement. Features are extracted from the EEG signals and both the neural network and ANFIS are used to classify the signals based on the features. The performance of the two classification approaches are then compared to determine if the hybrid ANFIS method can outperform the established neural network approach.
Classification of emotions induced by horror and relaxing movies using single-...IJECEIAES
It has been observed from recent studies that corticolimbic Theta rhythm from EEG recordings perceived as fear or threatening scene during neural processing of visual stimuli. In additions, neural oscillations’ patterns in Theta, Alpha and Beta sub-bands also play important role in brain’s emotional processing. Inspired from these findings, in this paper we attempt to classify two different emotional states by analyzing single-channel EEG recordings. A video clip that can evoke 3 different emotional states: neutral, relaxation and scary is shown to 19 college-aged subjects and they were asked to score their emotional outcome by giving a number between 0 to 10 (where 0 means not scary at all and 10 means the most scary). First, recorded EEG data were preprocessed by stationary wavelet transform (SWT) based artifact removal algorithm. Then power distribution in simultaneous time-frequency domain was analyzed using short-time Fourier transform (STFT) followed by calculating the average power during each 0.2s time-segment for each brain sub-band. Finally, 46 features, as the mean power of frequency bands between 4 and 50 Hz during each time-segment, containing 689 instances—for each subject —were collected for classification. We found that relaxation and fear emotions evoked during watching scary and relaxing movies can be classified with average classification rate of 94.208% using K-NN by applying methods and materials proposed in this paper. We also classified the dataset using SVM and we found out that K-NN classifier (when k = 1) outperforms SVM in classifying EEG dynamics induced by horror and relaxing movies, however, for K > 1 in K-NN, SVM has better average classification rate.
Overview of Machine Learning and Deep Learning Methods in Brain Computer Inte...IJCSES Journal
Research under the field of Brain Computer Interfaces is adapting various Machine Learning and Deep
Learning techniques in recent times. With the advent of modern BCI, the data generated by various devices
is now capable of detecting brain signals more accurately. This paper gives an overview of all the steps
involved in the process of applying Machine Learning as well as Deep Learning methods from Data
Acquisition to application of algorithms. It aims to study techniques currently employed to extract data,
features from brain data, different algorithms employed to draw insights from the extracted features, and
how it can be used in various BCI applications. By this study, I aim to put forward current Machine
Learning and Deep Learning Trends in the field of BCI.
International Journal of Engineering Research and Applications (IJERA) is an open access online peer reviewed international journal that publishes research and review articles in the fields of Computer Science, Neural Networks, Electrical Engineering, Software Engineering, Information Technology, Mechanical Engineering, Chemical Engineering, Plastic Engineering, Food Technology, Textile Engineering, Nano Technology & science, Power Electronics, Electronics & Communication Engineering, Computational mathematics, Image processing, Civil Engineering, Structural Engineering, Environmental Engineering, VLSI Testing & Low Power VLSI Design etc.
EEG S IGNAL Q UANTIFICATION B ASED ON M ODUL L EVELS sipij
This article proposes a contribution to quantify EE
G signals outline. This technique uses two tools fo
r EEG
signal characteristics extraction. Our tests were r
ealized on the basis of 32 canals EEG canals using
Neuroscan software. EEG example demonstration is re
ferenced CZ and is sampled at 1000HZ. The
principal aim of this technique is to reduce the im
portant volume of EEG signal data Without losing an
y
information. EEG signals are quantified on the basi
s of a whole predefined levels The obtained results
show that an EEG alignment can be posted in a quant
ified form.
A LOW COST EEG BASED BCI PROSTHETIC USING MOTOR IMAGERY ijitcs
Brain Computer Interfaces (BCI) provide the opportunity to control external devices using the brain
ElectroEncephaloGram (EEG) signals. In this paper we propose two software framework in order to
control a 5 degree of freedom robotic and prosthetic hand. Results are presented where an Emotiv
Cognitive Suite (i.e. the 1st framework) combined with an embedded software system (i.e. an open source
Arduino board) is able to control the hand through character input associated with the taught actions of
the suite. This system provides evidence of the feasibility of brain signals being a viable approach to
controlling the chosen prosthetic. Results are then presented in the second framework. This latter one
allowed for the training and classification of EEG signals for motor imagery tasks. When analysing the
system, clear visual representations of the performance and accuracy are presented in the results using a
confusion matrix, accuracy measurement and a feedback bar signifying signal strength. Experiments with
various acquisition datasets were carried out and with a critical evaluation of the results given. Finally
depending on the classification of the brain signal a Python script outputs the driving command to the
Arduino to control the prosthetic. The proposed architecture performs overall good results for the design
and implementation of economically convenient BCI and prosthesis.
A machine learning algorithm for classification of mental tasks.pdfPravinKshirsagar11
In this article, a contemporary tack of mental tasks on cognitive parts of humans is appraised using two different approaches such as wavelet transforms at a discrete time (DWT) and support vector machine (SVM). The put forth tack is instilled with the electroencephalogram (EEG) database acquired in real-time from CARE Hospital, Nagpur. Additional data is also acquired from a brain-computer interface (BCI). In the working model, signals from the database are wed out into different frequency sub-bands using DWT. Initially, updated statistical features are obtained from different frequency sub-bands. This type of representation defines the wavelet co-efficient which is introduced for reducing the measurement of data. Then, the projected method is realized using SVM for segregating both port and veracious hand movement. After segregation of EEG signals, results are achieved with an accuracy of 92% for BCI competition paradigm III and 97.89% for B-alert machine.
The document describes a brain-computer interface (BCI) system that uses electroencephalography (EEG) to classify motor imagery of the left or right arm and control an assistive device for paralyzed upper limbs. EEG signals are recorded over motor cortex areas during right and left arm imagery tasks. The mu and beta frequency bands are extracted and used to classify intended movement based on features like power and mean. If right arm imagery is classified, a stepper motor attached to the patient's forearm is activated to help lift their arm. The system was tested on 7 subjects with over 10 trials each, achieving classification of intended movement.
EEG SIGNAL CLASSIFICATION USING LDA AND MLP CLASSIFIERhiij
Electroencephalography (EEG) is the recording of electrical activities along the scalp. EEG measures
voltage fluctuations resulting from ionic current flows within the neurons of the brain. Diagnostic
applications generally focus on the spectral content of EEG, which is the type of neural oscillations that
can be observed in EEG signal. EEG is most often used to diagnose epilepsy, which causes obvious
abnormalities in EEG readings. This powerful property confirms the rich potential for EEG analysis and
motivates the need for advanced signal processing techniques to aid clinicians in their interpretations.
This paper describes the application of Wavelet Transform (WT) for the processing of
Electroencephalogram (EEG) signals. Furthermore, the linear discriminant analysis (LDA) is applied for
feature selection and dimensionality reduction where the informative and discriminative two-dimension
features are used as a benchmark for classification purposes through a Multi-Layers Perceptron (MLP)
neural network. For five classification problems, the proposed model achieves a high sensitivity,
specificity and accuracy of 100%.Finally, the comparison of the results obtained with the proposed
methods and those obtained with previous literature methods shows the superiority of our approach for
EEG signals classification and automated diagnosis
Sample size determination for classification of eeg signals using power analy...iaemedu
The document discusses determining the minimum sample size needed for classification of electroencephalogram (EEG) signals using machine learning. It proposes using power analysis to calculate the required sample size to separate classes with statistical stability. Power analysis was performed on a dataset of 500 EEG signals from 5 classes. The results found that a sample size of 81 signals is needed to achieve 95% power. Additional experiments varied the power level and error probability to relate their effects on minimum sample size. The sample sizes calculated from power analysis were validated using a decision tree classifier on the EEG dataset.
Similar to A Comparative Study of Machine Learning Algorithms for EEG Signal Classification (20)
We have compiled the most important slides from each speaker's presentation. This year’s compilation, available for free, captures the key insights and contributions shared during the DfMAy 2024 conference.
DEEP LEARNING FOR SMART GRID INTRUSION DETECTION: A HYBRID CNN-LSTM-BASED MODELgerogepatton
As digital technology becomes more deeply embedded in power systems, protecting the communication
networks of Smart Grids (SG) has emerged as a critical concern. Distributed Network Protocol 3 (DNP3)
represents a multi-tiered application layer protocol extensively utilized in Supervisory Control and Data
Acquisition (SCADA)-based smart grids to facilitate real-time data gathering and control functionalities.
Robust Intrusion Detection Systems (IDS) are necessary for early threat detection and mitigation because
of the interconnection of these networks, which makes them vulnerable to a variety of cyberattacks. To
solve this issue, this paper develops a hybrid Deep Learning (DL) model specifically designed for intrusion
detection in smart grids. The proposed approach is a combination of the Convolutional Neural Network
(CNN) and the Long-Short-Term Memory algorithms (LSTM). We employed a recent intrusion detection
dataset (DNP3), which focuses on unauthorized commands and Denial of Service (DoS) cyberattacks, to
train and test our model. The results of our experiments show that our CNN-LSTM method is much better
at finding smart grid intrusions than other deep learning algorithms used for classification. In addition,
our proposed approach improves accuracy, precision, recall, and F1 score, achieving a high detection
accuracy rate of 99.50%.
Low power architecture of logic gates using adiabatic techniquesnooriasukmaningtyas
The growing significance of portable systems to limit power consumption in ultra-large-scale-integration chips of very high density, has recently led to rapid and inventive progresses in low-power design. The most effective technique is adiabatic logic circuit design in energy-efficient hardware. This paper presents two adiabatic approaches for the design of low power circuits, modified positive feedback adiabatic logic (modified PFAL) and the other is direct current diode based positive feedback adiabatic logic (DC-DB PFAL). Logic gates are the preliminary components in any digital circuit design. By improving the performance of basic gates, one can improvise the whole system performance. In this paper proposed circuit design of the low power architecture of OR/NOR, AND/NAND, and XOR/XNOR gates are presented using the said approaches and their results are analyzed for powerdissipation, delay, power-delay-product and rise time and compared with the other adiabatic techniques along with the conventional complementary metal oxide semiconductor (CMOS) designs reported in the literature. It has been found that the designs with DC-DB PFAL technique outperform with the percentage improvement of 65% for NOR gate and 7% for NAND gate and 34% for XNOR gate over the modified PFAL techniques at 10 MHz respectively.
Literature Review Basics and Understanding Reference Management.pptxDr Ramhari Poudyal
Three-day training on academic research focuses on analytical tools at United Technical College, supported by the University Grant Commission, Nepal. 24-26 May 2024
CHINA’S GEO-ECONOMIC OUTREACH IN CENTRAL ASIAN COUNTRIES AND FUTURE PROSPECTjpsjournal1
The rivalry between prominent international actors for dominance over Central Asia's hydrocarbon
reserves and the ancient silk trade route, along with China's diplomatic endeavours in the area, has been
referred to as the "New Great Game." This research centres on the power struggle, considering
geopolitical, geostrategic, and geoeconomic variables. Topics including trade, political hegemony, oil
politics, and conventional and nontraditional security are all explored and explained by the researcher.
Using Mackinder's Heartland, Spykman Rimland, and Hegemonic Stability theories, examines China's role
in Central Asia. This study adheres to the empirical epistemological method and has taken care of
objectivity. This study analyze primary and secondary research documents critically to elaborate role of
china’s geo economic outreach in central Asian countries and its future prospect. China is thriving in trade,
pipeline politics, and winning states, according to this study, thanks to important instruments like the
Shanghai Cooperation Organisation and the Belt and Road Economic Initiative. According to this study,
China is seeing significant success in commerce, pipeline politics, and gaining influence on other
governments. This success may be attributed to the effective utilisation of key tools such as the Shanghai
Cooperation Organisation and the Belt and Road Economic Initiative.
International Conference on NLP, Artificial Intelligence, Machine Learning an...gerogepatton
International Conference on NLP, Artificial Intelligence, Machine Learning and Applications (NLAIM 2024) offers a premier global platform for exchanging insights and findings in the theory, methodology, and applications of NLP, Artificial Intelligence, Machine Learning, and their applications. The conference seeks substantial contributions across all key domains of NLP, Artificial Intelligence, Machine Learning, and their practical applications, aiming to foster both theoretical advancements and real-world implementations. With a focus on facilitating collaboration between researchers and practitioners from academia and industry, the conference serves as a nexus for sharing the latest developments in the field.
A SYSTEMATIC RISK ASSESSMENT APPROACH FOR SECURING THE SMART IRRIGATION SYSTEMSIJNSA Journal
The smart irrigation system represents an innovative approach to optimize water usage in agricultural and landscaping practices. The integration of cutting-edge technologies, including sensors, actuators, and data analysis, empowers this system to provide accurate monitoring and control of irrigation processes by leveraging real-time environmental conditions. The main objective of a smart irrigation system is to optimize water efficiency, minimize expenses, and foster the adoption of sustainable water management methods. This paper conducts a systematic risk assessment by exploring the key components/assets and their functionalities in the smart irrigation system. The crucial role of sensors in gathering data on soil moisture, weather patterns, and plant well-being is emphasized in this system. These sensors enable intelligent decision-making in irrigation scheduling and water distribution, leading to enhanced water efficiency and sustainable water management practices. Actuators enable automated control of irrigation devices, ensuring precise and targeted water delivery to plants. Additionally, the paper addresses the potential threat and vulnerabilities associated with smart irrigation systems. It discusses limitations of the system, such as power constraints and computational capabilities, and calculates the potential security risks. The paper suggests possible risk treatment methods for effective secure system operation. In conclusion, the paper emphasizes the significant benefits of implementing smart irrigation systems, including improved water conservation, increased crop yield, and reduced environmental impact. Additionally, based on the security analysis conducted, the paper recommends the implementation of countermeasures and security approaches to address vulnerabilities and ensure the integrity and reliability of the system. By incorporating these measures, smart irrigation technology can revolutionize water management practices in agriculture, promoting sustainability, resource efficiency, and safeguarding against potential security threats.
Introduction- e - waste – definition - sources of e-waste– hazardous substances in e-waste - effects of e-waste on environment and human health- need for e-waste management– e-waste handling rules - waste minimization techniques for managing e-waste – recycling of e-waste - disposal treatment methods of e- waste – mechanism of extraction of precious metal from leaching solution-global Scenario of E-waste – E-waste in India- case studies.
A review on techniques and modelling methodologies used for checking electrom...nooriasukmaningtyas
The proper function of the integrated circuit (IC) in an inhibiting electromagnetic environment has always been a serious concern throughout the decades of revolution in the world of electronics, from disjunct devices to today’s integrated circuit technology, where billions of transistors are combined on a single chip. The automotive industry and smart vehicles in particular, are confronting design issues such as being prone to electromagnetic interference (EMI). Electronic control devices calculate incorrect outputs because of EMI and sensors give misleading values which can prove fatal in case of automotives. In this paper, the authors have non exhaustively tried to review research work concerned with the investigation of EMI in ICs and prediction of this EMI using various modelling methodologies and measurement setups.
ACEP Magazine edition 4th launched on 05.06.2024Rahul
This document provides information about the third edition of the magazine "Sthapatya" published by the Association of Civil Engineers (Practicing) Aurangabad. It includes messages from current and past presidents of ACEP, memories and photos from past ACEP events, information on life time achievement awards given by ACEP, and a technical article on concrete maintenance, repairs and strengthening. The document highlights activities of ACEP and provides a technical educational article for members.
Embedded machine learning-based road conditions and driving behavior monitoringIJECEIAES
Car accident rates have increased in recent years, resulting in losses in human lives, properties, and other financial costs. An embedded machine learning-based system is developed to address this critical issue. The system can monitor road conditions, detect driving patterns, and identify aggressive driving behaviors. The system is based on neural networks trained on a comprehensive dataset of driving events, driving styles, and road conditions. The system effectively detects potential risks and helps mitigate the frequency and impact of accidents. The primary goal is to ensure the safety of drivers and vehicles. Collecting data involved gathering information on three key road events: normal street and normal drive, speed bumps, circular yellow speed bumps, and three aggressive driving actions: sudden start, sudden stop, and sudden entry. The gathered data is processed and analyzed using a machine learning system designed for limited power and memory devices. The developed system resulted in 91.9% accuracy, 93.6% precision, and 92% recall. The achieved inference time on an Arduino Nano 33 BLE Sense with a 32-bit CPU running at 64 MHz is 34 ms and requires 2.6 kB peak RAM and 139.9 kB program flash memory, making it suitable for resource-constrained embedded systems.
Electric vehicle and photovoltaic advanced roles in enhancing the financial p...IJECEIAES
Climate change's impact on the planet forced the United Nations and governments to promote green energies and electric transportation. The deployments of photovoltaic (PV) and electric vehicle (EV) systems gained stronger momentum due to their numerous advantages over fossil fuel types. The advantages go beyond sustainability to reach financial support and stability. The work in this paper introduces the hybrid system between PV and EV to support industrial and commercial plants. This paper covers the theoretical framework of the proposed hybrid system including the required equation to complete the cost analysis when PV and EV are present. In addition, the proposed design diagram which sets the priorities and requirements of the system is presented. The proposed approach allows setup to advance their power stability, especially during power outages. The presented information supports researchers and plant owners to complete the necessary analysis while promoting the deployment of clean energy. The result of a case study that represents a dairy milk farmer supports the theoretical works and highlights its advanced benefits to existing plants. The short return on investment of the proposed approach supports the paper's novelty approach for the sustainable electrical system. In addition, the proposed system allows for an isolated power setup without the need for a transmission line which enhances the safety of the electrical network
Electric vehicle and photovoltaic advanced roles in enhancing the financial p...
A Comparative Study of Machine Learning Algorithms for EEG Signal Classification
1. Signal & Image Processing: An International Journal (SIPIJ) Vol.12, No.6, December 2021
DOI: 10.5121/sipij.2021.12603 37
A COMPARATIVE STUDY OF MACHINE LEARNING
ALGORITHMS FOR EEG SIGNAL CLASSIFICATION
Anam Hashmi, Bilal Alam Khan and Omar Farooq
Department of Electronics Engineering, Aligarh Muslim University, Aligarh, India
ABSTRACT
In this paper, different machine learning algorithms such as Linear Discriminant Analysis, Support vector
machine (SVM), Multi-layer perceptron, Random forest, K-nearest neighbour, and Autoencoder with SVM
have been compared. This comparison was conducted to seek a robust method that would produce good
classification accuracy. To this end, a robust method of classifying raw Electroencephalography (EEG)
signals associated with imagined movement of the right hand and relaxation state, namely Autoencoder
with SVM has been proposed. The EEG dataset used in this research was created by the University of
Tubingen, Germany. The best classification accuracy achieved was 70.4% with SVM through feature
engineering. However, our prosed method of autoencoder in combination with SVM produced a similar
accuracy of 65% without using any feature engineering technique. This research shows that this system of
classification of motor movements can be used in a Brain-Computer Interface system (BCI) to mentally
control a robotic device or an exoskeleton.
KEYWORDS
EEG. Machine learning. BCI. Motor Imagery signals. Random Forest.
1. INTRODUCTION
Assistive technologies have witnessed tremendous attention and advancements both from the
scientific community and industry partners in the last couple of decades. This has led to
significant innovation and improvements in the following sector and fields: virtual surgical
theatre, robotic surgery, Brain-controlled wheelchairs are the name of the few recent
developments. The field of brain-computer interface has also caught the attention of researchers
from different fields including neuroscience, cognitive psychology, computer science, and
electrical engineering – as it provides the avenue for human welfare and improving life
experience. It can be observed inefficient disease diagnosis, development of assistive
technologies, health monitoring of the elderly and aiding humanity in general [1]. This study also
seeks to explore further this very dimension by analyzing different methodologies used in
studying Brain-Computer Interface (BCI). Electroencephalogram or EEG is one of the most
common non-invasive methodologies of BCI to record brain signals. It measures the electrical
activity of the brain using electrodes that are placed over the scalp. EEG is preferred because of
its ease of portability and capturing high temporal brain information, however, it fails in
capturing high spatial information [2]. BCI uses these EEG signals associated with the user’s
activity and then apply different signal processing algorithms for translating the recorded signals
into control commands for different applications. In an EEG there are five types of oscillatory
waves that are commonly used for analysis, which are:
(a) delta (0.5–4 Hz);
(b) theta (4–7 Hz);
2. Signal & Image Processing: An International Journal (SIPIJ) Vol.12, No.6, December 2021
38
(c) alphaormu (7–13 Hz);
(d) beta (13–25 Hz);
(e) gamma (25–50 Hz).
Motor imagery (MI) is a process in which an individual rehearses or stimulates an action. It is a
very popular paradigm in the analysis of an EEG based BCI system. MI activity usually lies in
alpha (or mu) and beta bands [3].
In the past few years, significant advances have been made in the BCI systems and they have
revolutionized rehabilitation engineering by providing differently-abled individuals with a new
avenue to communicate with the external environment. According to many works of literature,
the strength of a BCI system depends upon the methods in which the brain signals are translated
into control commands of machines. A novel method namely an arc detection algorithm to find
an optimal channel was proposed by ErdemEkran and Ismail Kurnaz [4]. For feature extraction
DWT was used and several machine learning algorithms were used for classification purposes,
which were SVM, K- nearest neighbour, and Linear Discriminant Analysis. The best accuracy
achieved by their methodology was 95% in classifying ECoG signals (BCI competition III,
dataset I). Jun Wang and Yan Zhao proposed feature selection based on one dimension real-
valued particle swarm optimization, extracted nonlinear features such as Approximate entropy
and Wavelet packet decomposition, and achieved the best accuracy of 100% [5]. Khan B. A. et
al. (2020), employed feature engineering and linear discriminant analysis to accurately classify
EEG signals associated with the seizure. The authors have suggested the use of novel features
such as Gini’s coefficient to extract features and employed a very simple Linear discriminant
analysis based method to accurately classify seizure data with an accuracy of 100% [16]. In
another work, the authors' Khan B. A. et al., have used very simple statistical features to classify
imagined and executed hand movements using EEG signals [17]. Aswinseshadri. K et al. used the
wavelet packet tree for feature extraction. They used a genetic algorithm, applied information
gain, and mutual information to find the best feature set and for classification K-NN and Naïve
Bayes were employed [6]. Chea-Yau Kee et al. proposed a novel feature known as Renyi entropy
that has been employed for feature extraction and BLDA for classification [7]. K. Venkatachalam
et al. proposed the use of the Hybrid-KELM (Kernel Extreme Learning Machine) method based
on PCA (Principal Component Analysis) and FLD (Ficher’s Linear Discriminant) analysis for MI
BCI classification of EEG signals. The best accuracy reported was 96.54% [8]. Rajdeep
Chatterjee et al. used the AAR (Auto Adaptive Regressive) algorithm for feature extraction,
proposed a novel feature selection method based on IoMT (Internet of Medical Things), and
classified EEG signals using SVM and ensemble variants of classifiers. The best accuracy
reported was 80% [9]. The authors of [10] employed a combination of common spatial patterns
(CSP) and local characteristic-scale decomposition (LCD) algorithm for feature extraction, a
combination of firefly algorithm and learning automata (LA) to optimize feature selection, and
spectral regression discriminant analysis (SRDA) classifier for classifying MI-EEG signals. They
have used this method for a real-time brain-computer interface to show their method’s efficiency.
Several studies have usually worked on the classification of right vs left-hand movement, or hand
vs tongue movements, or hands vs legs movements. There is very limited literature that has
studied and classified intricate hand movements such as opening and closing of a hand, or
movements of different fingers, or classification of different hand gestures using neural signals,
and those who have worked on these subjects either did not achieve high enough accuracy or
failed to work in a real-world setup. This paper probed this very aspect of studying intricate
human motions and worked on the classification of imagining of opening and relaxing of a hand
using MI-EEG signals.
3. Signal & Image Processing: An International Journal (SIPIJ) Vol.12, No.6, December 2021
39
The contributions of this paper are following:
(i) Comparison of different machine learning algorithms to empirically establish a proper
method that could provide satisfactory results for this dataset.
(ii) Accurate classification of the motor imagery signals using a deep learning-based
algorithm (Autoencoder) which utilizes raw EEG signals and sends them for
classification. Since the pipeline is made independent of this dataset, it should work for
other similar physiological signal datasets.
The organization of this paper is as follows: the first part is the Introduction stage, where a brief
introduction was provided and related works were reported, followed by the Materials and
Methodology stage. In this part, the materials or data that was used in the paper is described and
the methodology of this work was elucidated. The third stage involves the results of the study
with a detailed discussion and a conclusion along with the limitation of this study and future
scope.
2. MATERIALS AND METHODOLOGY
2.1. Data Used
The data used in this study was taken from [11]. The data consist of EEG recordings of a single
subject. The subject was connected with a high spinal cord lesion and was controlling an
exoskeleton (Brain-Neural computer interface) attached to his paralysed limb. The cue-based
BNCI paradigm consisted of two different tasks, namely the ‘imagination of movement’ of the
right hand (Class 1) and ‘relaxation/no movement’ (Class 2).
Figure 1. Timing Scheme of the trials used (a) and a subject during the EEG recording of the Dataset (b).
A randomly shown visual cue is used to indicate to the user when to open (for Green square) and
when to close (for Red Square). These two indications were given 24 times each in total
separated by inter-trial intervals (ITIs) of 4-6 seconds. Each indication was displayed for 5
seconds after which the device was driven back to the open position. Re-setting the exoskeleton
into open position required one second.
EEG was recorded from 5 conventional EEG recording sites F4, T8, C4, Cz, and P4 according to
the international 10/20 system using an active electrode EEG system (Acti-cap® and
BrainAmp®, BrainProducts GmbH, Gilching, Germany) with a reference electrode placed at FCz
and ground electrode at AFz. EEG was recorded at a sampling rate of 200 Hz, bandpass filtered
at 0.4-70Hz and pre-processed using a small Laplacian filter.
4. Signal & Image Processing: An International Journal (SIPIJ) Vol.12, No.6, December 2021
40
Figure 2. Showing raw EEG channel waveforms of different channels [11].
2.2. Pre-Processing
At this stage, the data was processed or filtered to capture information related to Motor Imagery.
While the recording of EEG stores different noise elements from line frequencyinterference to
different unwanted artefact signals. All of which could frustrate the model’s classification
progress. Therefore, it is necessary to first remove these noise elements and unwanted signals
before the actual analysis. Many electrophysiological features are associated with the brain’s
normal motor output channels [12]. Several studies have suggested the presence of mu rhythm in
the frequency range of 7-13 Hz. Some of these important features are the mu (8-12 Hz) and beta
(13-30 Hz) rhythms [13]. Therefore, this particular band has been selected for further analysis. To
this end, an FIR filter was designed for the frequency range of 6 to 14 Hz with Hamming window
of 0.0194 passband ripple and 53 dB stopband attenuation.
(a)
5. Signal & Image Processing: An International Journal (SIPIJ) Vol.12, No.6, December 2021
41
(b)
Figure 3. Showing power spectral density of EEG signal channel ‘C4’ (a) before filtering and (b) after
filtering using the designed FIR filter.
2.3. Feature Extraction
A feature is a measurable property or characteristic of an observed signal. It should be
informative, discriminative and orthogonal to other features. Feature extraction is the method of
extracting these features. It can be defined as the process of transforming original data into a
dataset with a reduced number of variables but with the most discriminative information.
After the pre-processing stage, the channels were selected from the F4, T8, C4, Cz and P4 based
on literature review and using correlation-based analysis. Finally, Cz and P4 channels were
selected as they were amongst the most commonly used channels for motor imager classification
purposes [17]. To compare different machine learning algorithms certain features were extracted.
The choice of these features was based on the results from our previous work [19] and other
commonly used features. These features include Mean absolute value (MAV), Variance, Median
Absolute Deviation (MAD), Variance, Energy, Spectral Entropy, and Mean. Particularly, the two
classes – which corresponds to imagining of the opening of hand as ‘class 1’ and relaxation or no
movement as ‘class 2’ – differ in dispersion. The same can be observed from the histogram of
class 1 and class 2 appears, where the class 1 histogram appears to be skewed from the normal
distribution. Thereby justifying the choice of IQR, MAD, Variance, Standard Deviation,
Skewness and Kurtosis. Energy and MAV were chosen because it has been reported in many
works that mu rhythm has a lower amplitude than that of the alpha wave [14].
The following are the mathematical equations of the extracted features:
2.3.1. Mean Absolute Value (MAV)
It is defined as the mean value of the absolute values of the data. Mathematically,
𝑀𝐴𝑉 =
1
𝑁
∑ |𝑋𝑖(𝑛)|
𝑁
𝑖=1 (i)
2.3.2. Variance
It is defined as the expectations of the squared deviation of a random variable from its mean.
𝑉𝑎𝑟(𝑋) = 𝐸[(𝑋 − 𝜇)2] (ii)
6. Signal & Image Processing: An International Journal (SIPIJ) Vol.12, No.6, December 2021
42
Where Var(X) computes of variance of data X, ‘µ’ represents the average value, ‘E’ represents
the expectation.
2.3.3. Median Absolute Deviation
It is defined, as the name suggests, as the median value of the absolute deviations from the data
median value.
𝑀𝐴𝐷 = 𝑚𝑒𝑑𝑖𝑎𝑛(|𝑋𝑖 − 𝑚𝑒𝑑𝑖𝑎𝑛(𝑋)|) (iii)
Where Xi is the ith value of the data X.
2.3.4. Spectral Entropy
The spectral entropy (SE) of a signal is a measure of its spectral power distribution. X(m) is the
discrete Fourier transform of the signal x(n). S(m) is the power spectrum of the X(m). P(m) is the
probability distribution of S(m) and H is the Spectral Entropy, calculated based on the Shannon
entropy formula.
𝑆(𝑚) = |𝑋(𝑚)2
|
𝑃(𝑚) =
𝑆(𝑚)
∑ 𝑆(𝑖)
𝑖
𝐻 = − ∑ 𝑃(𝑚)𝑙𝑜𝑔2𝑃(𝑚)
𝑁
𝑚=1 (iv)
2.3.5. Energy
It is the area under the squared magnitude of the considered signal. Mathematically,
𝐸𝑠 = ∑ |𝑋(𝑛)|2
∞
𝑛= −∞ (v)
2.3.6. Mean
It is defined as the averaged sum of a series of numbers. It can be calculated as,
𝑀𝑒𝑎𝑛 =
∑ 𝑋𝑖
𝑖
𝑛
(vi)
2.4. Methodology
The study followed a very simple pipeline –starting from the pre-processing of unwanted
artefacts and channel selection, then feature were extracted, and finally, different combinations of
features were classified using different machine learning algorithms to empirically observe which
algorithm would work best for this task.
In this study, two classes are corresponding to the Motor Imagery (MI) tasks; hand opens and
hand relaxes. The data was pre-processed and filtered using an FIR bandpass filter. This band
was considered as it corresponds to the Mu rhythm (7 – 13 Hz) where the motor activity in the
brain is usually associated [khan B. A.]. The total duration of ‘Class 1’ is 45 seconds and the
sample rate is 200 Hz, which produces 9000 data points.
One of the pipelines used in this study involves the use of traditional feature engineering steps
and employing a classifier to do classification. To this end, the following steps have been
7. Signal & Image Processing: An International Journal (SIPIJ) Vol.12, No.6, December 2021
43
followed. A one-second sliding window is considered for the analysis of the signal and each
second 200 samples are considered, for which 6 features were extracted which produce a feature
vector of [45×1] corresponding to each feature. This process was performed until the end of the
recording, thereby producing a feature vector of the size of [270×1]. This was repeated for the
selected channels (Cz and P4) producing a feature matrix of [270×2]. After the feature extraction,
different combinations of these features were considered. At each selected feature combination
different training and testing sizes were split to see classification performance across different
training and test split sizes. This has also provided insight into algorithms that are relatively more
robust to such changes. Table 1. Shows different feature combinations considered in this study.
The other pipeline – based on autoencoder with other classifiers – employed in this study does
not make use of traditional feature engineering steps and skips over this. It directly utilizes raw
EEG signals followed by the classification stage using SVM, Random Forest, and K-nearest
neighbour. Details about these classifiers and used methodology have been described below:
Table 1. showing features and their combination.
Feature name Feature code Feature combination
Mean absolute value F1 F5
Variance F2 F5+F6
Median absolute deviation F3 F3+F5+F6
Spectral Entropy F4 F1+F3+F5+F6
Energy F5 F1+F2+F3+F5+F6
Mean F6 F1+F2+F3+F4+F5+F6
Here, Autoencoder will be described as the other methodology is defined above. Autoencoders
are feedforward neural networks that can learn efficient representations of the input data without
the need for labels in the training data. Autoencoders are regarded as powerful feature extractors.
So, autoencoders work by learning efficient ways to represent the input data by copying their
inputs to their outputs. In the learning process of the autoencoders, we can put several constraints
on the way these learn the internal representations of the input data, such as reducing the number
of features, which will make the autoencoder work as dimensionality reduction networks. The
architecture of the autoencoder we used consisted of one input layer, one hidden layer and one
output layer. The number of nodes used in the hidden layer was 5. The autoencoder network
architecture consists of an encoder part and a decoder part, the encoder part is responsible for
coding the input data while the decoder part reconstructs the input from the code [18]. The
activation function we used in the autoencoder network was Exponential Linear Unit (ELU). The
reason for choosing the ELU activation function was that it is a non-saturating function and thus,
doesn’t suffer from the vanishing gradients problem. After encoding and decoding raw EEG
signals, the decoded signals are sent for classification using the best classifiers observed by using
classification accuracy as the metrics. These classifiers included – SVM, Random forest, and K-
nearest neighbour using Sklearn library of python [20]. Their corresponding accuracies are
reported in the result section.
8. Signal & Image Processing: An International Journal (SIPIJ) Vol.12, No.6, December 2021
44
Figure 4. Autoencoder depiction with input layer, hidden layer and output layer. In this study, input layer
has the dimension of input data, hidden layer has only 5 nodes and output layer is again of the same size as
input.
K-nearest neighbour (KNN)
The k-nearest neighbours (KNN) algorithm is a non-parametric, supervised machine learning
algorithm, that is both simple and powerful. The KNN algorithm works by assuming that the data
points that are similar exists in close proximity to each other and works on the idea of a similarity
function (example, distance functions like Euclidean).
This method is used independently as well as along with the autoencoder pipeline. For the KNN
classifier, the values of hyperparameters chosen were: the number of neighbours to use
(n_neighbors) was taken as 5 and the distance metric we chose was ‘Euclidean’ distance.
Support Vector Machine (SVM)
One of the most powerful and versatile machine learning algorithms is Support Vector Machine
(SVM), which can perform both linear and non-linear classification. The Support vector
machines work by finding a hyperplane which is a decision boundary in N-dimensional space (N-
the dimensions of feature space) for distinctly classifying the data points. The objective of the
SVM is to generate a maximal marginal hyperplane that can divide the dataset into distinct
classes in the best possible way. Next, we used Support Vector Machines using the Sklearn
library of python [20].
For the SVM classifier, the values of hyperparameters chosen were 0.1 for the regularization
parameter C, the kernel used was ‘radial basis function’ (‘rbf’) and the value for the kernel
coefficient ‘gamma’ used was 1.
Random Forest Classifier
Random Forests are known as ensemble learning classifiers and usually gives good results
without much hyperparameter tuning. These work by constructing a number of decision trees
during training by using the split criteria for the decision nodes as ‘Gini’ impurity or ‘entropy,
9. Signal & Image Processing: An International Journal (SIPIJ) Vol.12, No.6, December 2021
45
and the output of which is chosen as the class most selected by the decision trees.Finally, we used
the Random Forest Classifier using the Sklearn library of python [20].
For the Random Forest classifier, the values of hyperparameters chosen were 300 for the number
of trees in the forest (n_estimators), the function to measure the quality of split of the nodes of the
decision was ‘Gini’ impurity.
Multi-layer Perceptron (MLP)
The perceptron is a single neuron while the Multi-layer Perceptron (MLP) is a class of artificial
neural networks that uses supervised learning and are composed of multiple layers of the
perceptron. Multi-layer perceptron can classify data that is not linearly separable unlike that of a
perceptron and consists of at least three layers, the input layer, the hidden layer, and an output
layer.
Linear Discriminant Classifier (LDA)
Linear Discriminant Classifier (LDA) is a very simple supervised classification algorithm that
works by finding a combination of linear features to separate the data into two or more classes.
LDA considers two assumptions for the classification task, one is that it assumes that the data has
Gaussian distribution and the second is that the classes in the dataset have the same covariance
matrices.
3. CLASSIFICATION
The classification of EEG signals plays a vital role in biomedical research. According to [15],
there are mainly 5 types of classifiers used in BCI research such as linear classifiers,
nonlinearclassifiers, neural networks, nearest neighbour classifiers and a combination of these. In
this study, all of these classifiers have been compared to empirically establish which classifier
would be most appropriate for this “task”. These include Linear Discriminant Analysis (LDA),
Support Vector Machine (SVM), K-Nearest Neighbour (KNN), Random Forest (RF), Multi-layer
Perceptron (MLP) and the combination of these classifiers using majority voting criteria. The
best three classifiers were selected based on the classification accuracy to accurately classify
motor imagery signals.
Finally, a deep learning-based autoencoder along with SVM was used to classify imagined
movement of hand using different classifiers including SVM, Random Forest, and K-nearest
neighbour. EEG signals were directly input to the autoencoder which encodes and decode raw
EEG and performs dimensional reduction. The decoded output will be used in the classification
stage and will be classified using these different mentioned classifiers.
4. RESULTS
Different machine learning algorithms have been used in this study by comparing their
classification accuracy and their robustness by changing parameters such as feature combination
and train-test data size. Results from these changes have been reported here below from Table 2
to Table10. We have used classification accuracy in order to evaluate the effectiveness of our
method.
𝐶𝑙𝑎𝑠𝑠𝑖𝑓𝑖𝑐𝑎𝑡𝑖𝑜𝑛 𝑎𝑐𝑐𝑢𝑟𝑎𝑐𝑦 (%) =
𝑇𝑃+𝑇𝑁
𝑇𝑃+𝐹𝑃+𝐹𝑁+𝑇𝑁
(i)
Where,
10. Signal & Image Processing: An International Journal (SIPIJ) Vol.12, No.6, December 2021
46
TP is True Positive;
TN is True Negative;
FP is False Positive;
FN is False Negative
Table 2 summarises results from Table 3 to Table 10. It provides a brief overview of how
different classifiers have been used and which provided the best accuracy. Moreover, it also
shows the combination of classifiers that have provided the best accuracy based on majority
voting criteria. Table 2a shows a different combination of classifiers. For combining, classifiers
are so chosen based on their classification accuracy. The top three classifiers have been chosen
for each feature combination. Finally, only that classifier combination is reported which yielded
the best performance. Table 3 onwards reports results for classification using different feature
combinations – as has already been shown in Table 1 – by different classifiers and by their
combination as well. As can be seen in Table 3, shows results for different feature combinations
and corresponding classifier accuracies. Based on this table, the three best classifier combinations
are selected and whichever combination yields higher accuracy were reported. These best
combination accuracies have been reported in Table 4.
Table 2. Summary table of best classification accuracy corresponding to different feature combination.
S.no Feature combination Best train-
test split
Best
classifiers
Best combination accuracy
1 F1+F2+F3+F4+F5+F6 85-15 SVM - 70.4 63 (RF+SVM+KNN)
2 F1+F2+F3+F5+F6 85-15 SVM- 66.67 66.67 (RF+SVM+KNN)
3 F1+F3+F5+F6 85-15 SVM- 66.67 63 (RF+SVM+KNN)
4 F3+F5+F6 85-15 SVM- 70.4 63 (MLP+SVM+KNN)
5 F5+F6 85-15 MLP - 66.67 63 (MLP+SVM+KNN)
6 F5 85-15 SVM- 70.4 70.4 (MLP+SVM+KNN)
Table 2a. Showing different combination of classifiers utilizing majority voting criterion for improving
their performance.
S. No. Combination of Classifier Acronym
1 Random Forest + Support Vector Machine + K-nearest neighbour :
(RF+SVM+KNN)
T1
2 Multi-layer Perceptron + Support Vector Machine + K-nearest
neighbour : (MLP+SVM+KNN)
T2
3 Linear Discriminant Analysis+ Random Forest + Support Vector
Machine + Multi-layer Perceptron + K-nearest neighbour
(LDA+RF+SVM+MLP+KNN)
T3
Table 3. Classification accuracies of different classifiers for channels ‘Cz’ and ‘P4’ for different feature
combination.
S. No. Train-
Test
split
Features LDA RF MLP KNN SVM
1 85-15 F1+F2+F3+F4+F5+F6 44.45 55.6 44.45 55.6 70.4
2 85-15 F1+F2+F3+F5+F6 51.9 63 44.45 55.56 66.67
3 85-15 F1+F3+F5+F6 48.14 51.8 44.45 44.45 66.67
4 85-15 F3+F5+F6 40.7 48.14 59.25 51.85 70.37
5 85-15 F5+F6 44.45 55.56 66.67 62.9 62.9
6 85-15 F5 29.6 48.14 66.67 59.25 70.37
11. Signal & Image Processing: An International Journal (SIPIJ) Vol.12, No.6, December 2021
47
Table 4. Showing best classification accuracy by combining best three classifiers using majority criterion
for different combination of features.
Train-Test
split
Features Best three classifiers Best Classification
accuracy using
majority voting
criteria
85-15 F1+F2+F3+F4+F5+F6 T1: MLP+SVM+KNN 62.96
85-15 F1+F2+F3+F5+F6 T2: RF+SVM+KNN 66.7
85-15 F1+F3+F5+F6 RF+SVM+KNN 62.96
85-15 F3+F5+F6 MLP+SVM+KNN 62.96
85-15 F5+F6 MLP+SVM+KNN 62.96
85-15 F5 MLP+SVM+KNN 70.37
Table 5. Showing results from a six selected combination of features for different training and test data
split.
Table 6. Showing results from a five selected combination of features for different training and test data
split.
Train-
Test
Split
Feature
combination
selected
LDA RF MLP SVM KNN Best
combined
80-20 F1+F2+F3+F5+F6 47.22 58.33 55.56 55.56 55.56 63.88 (T1)
70-30 F1+F2+F3+F5+F6 51.85 53.7 44.44 57.4 53.7 55.56 (T1)
60-40 F1+F2+F3+F5+F6 54.16 52.8 40.3 54.2 55.56 55.56 (T3)
50-50 F1+F2+F3+F5+F6 51.11 53.33 47.8 57.8 60 58.9 (T1)
Table 7. Showing results from a four selected combination of features for different training and test data
split.
Train-
Test Split
Feature
combination
selected
LDA RF MLP SVM KNN Best
combined
80-20 F1+F3+F5+F6 44.44 55.56 50 55.56 47.22 58.33 (T1)
70-30 F1+F3+F5+F6 46.9 57.4 44.44 61.11 51.85 59.3 (T2)
60-40 F1+F3+F5+F6 47.22 52.78 40.3 56.9 51.3 52.8 (T1)
50-50 F1+F3+F5+F6 45.56 51.11 47.8 55.56 52.22 52.2 (T1)
Table 8. Showing results from a three selected combination of features for different training and test data
split.
Train-
Test Split
Feature
combination
selected
LDA RF MLP SVM KNN Best
combined
80-20 F3+F5+F6 50.0 55.56 66.67 51.11 55.56 61.11 (T2)
70-30 F3+F5+F6 42.59 48.14 44.44 68.52 55.56 61.11 (T2)
Train-
Test
Split
Feature combination
selected
LDA RF MLP SVM KNN Best combined
80-20 F1+F2+F3+F4+F5+F6 41.67 61.11 47.22 55.56 58.33 61.11 (T1)
70-30 F1+F2+F3+F4+F5+F6 46.3 55.56 44.44 50 55.56 57.4 (T3)
60-40 F1+F2+F3+F4+F5+F6 50 56.9 40.3 40.3 47.22 47.22 (T1)
50-50 F1+F2+F3+F4+F5+F6 46.7 52.22 47.8 57.8 47.8 50 (T1)
12. Signal & Image Processing: An International Journal (SIPIJ) Vol.12, No.6, December 2021
48
60-40 F3+F5+F6 41.67 51.4 40.3 66.67 55.56 61.11 (T2)
50-50 F3+F5+F6 40 56.67 47.78 61.11 55.56 58.89 (T1)
Table 9. Showing results from a two selected combination of features for different training and test data
split.
Train-
Test Split
Feature
combination
selected
LDA RF MLP SVM KNN Best
combined
80-20 F5+F6 44.44 58.33 63.88 58.33 55.56 58.23 (T1)
70-30 F5+F6 42.22 48.89 62.22 57.80 55.6 62.22 (T2)
60-40 F5+F6 36.11 51.38 45.83 48.61 47.22 50.0 (T1)
50-50 F5+F6 41.11 57.78 41.12 56.67 54.45 60.0 (T1)
Table 10. Showing results from a single for different training and test data split.
Train-
Test Split
Feature
combination
selected
LDA RF MLP SVM KNN Best
combined
80-20 F5 30.55 52.78 66.67 66.67 58.33 66.67 (T2)
70-30 F5 29.62 46.29 55.6 53.7 51.8 54.67 (T2)
60-40 F5 26.38 47.22 36.11 48.61 47.22 47.22 (T1)
50-50 F5 32.23 53.34 52.22 53.33 56.67 57.78 (T1)
Table 11. showing results from Autoencoder based method of classification.
S. NO. FEATURE CLASSIFIER ACCURACY (%)
1 AUTOENCODER KNN 58
2 AUTOENCODER SVM 65
3 AUTOENCODER RANDOM FOREST 63
5. DISCUSSION AND FUTURE WORK
In this study, different statistical features such as Mean Absolute Value, Median Absolute
Deviation, Variance, Spectral Entropy, Mean, and Energy were used to extract the underlying
information from a dynamic EEG. This study has shown that the proposed features were
successful in capturing the relevant distinguishing information. This study has also compared
different machine learning algorithms with one another under different conditions to observe the
method’s robustness to small changes. From the results, it can be observed that SVM has
produced consistently decent accuracy and was very least affected by changes in different feature
combinations or changes in train-test data split. After SVM, KNN and Random Forest were two
other algorithms that have shown significant promise in that aspect. On the other hand, Linear
Discriminant analysis has consistently performed very poorly. This is something that was
expected as the data is not linearly separable, which means linear classifiers would not be able to
work well on the dataset. This would also explain the inconsistent performance of MLP. This can
also be observed from the best classifier combination – most of the classifier combination does
not include a linear classifier. The summary table accurately summarizes this aspect in showing
the classifier and method that have been robust and also produced decent classification
accuracies.The important thing to note here is that the SVM has been reported in this study as the
robust method for the classification of EEG signals. However, this does not mean that its
performance was indifferent to the changes in the train-test data split. Its performance does get
affected – it decreases with increasing train-test data split – but is relatively stable when
13. Signal & Image Processing: An International Journal (SIPIJ) Vol.12, No.6, December 2021
49
compared with other classifiers. The best accuracy has also been achieved with SVM based
method of 70%. In our previous work, Hashmi A. et al. 2021, the best accuracy of 85% was
achieved after carefully cleaning the signals, applying pre-processing steps, extracting desired
frequency range through extremely computationally expensive method – wavelet decomposition
method, carefullyextracting relevant features and then ranking those features with the help of
random forest after which classification process was performed [19]. In this study, we tried to
reduce these computational expensive methods and replaced them with other very simple
methods such as using a simple FIR bandpass filter instead of wavelet decomposition to gain the
desired frequency range. Even after so many additional processing steps, the overall average
accuracy was around 75% which is quite similar to what this study has achieved. Also, this study
has developed another method – autoencoder with SVM. In this method, raw EEG signals will be
directly input in the autoencoder which will encode and decode it – reducing the dimensionality.
The decoded output will go into the SVM classifier which has been classified into two classes.
The average accuracy in this method was 65% and this method was very much indifferent to any
type of changes in the train-test data split. This method was developed without considering any of
the specificities of this dataset, so theoretically this method should be able to work decently on
other physiological datasets. The other classifier used with the autoencoder method were K-
nearest neighbour and Random forest. The corresponding accuracies for these methods were 58%
and 63%. It is worthwhile to mention here that these accuracies were not affected at all by the
change in train-test data split, meaning this method is quite robust for future use. However, there
is a need to improve these accuracy results by doing some sort of pre-processing of data. This can
be further explored in future studies – how much pre-processing would be necessary to make a
significant increase in classification accuracy? The other limitation of this study is that the data,
although very real-world and relevant, was very limited. This can be frustrating if deep learning-
based methods are to be employed. With these limitations, this study has tried to empirically
observe the effect of certain parameters such as train-test data split and features on the
classification accuracy and has compared different machine learning-based algorithms against
each other. A new robust approach based on deep learning-based methods has also been proposed
along with the comparison with our previous work.
The application of this method in the future is that it can be used to control an external device i.e.
Neuro-prosthetics. The translated commands will be used as input to the external device via a
computer (or micro-controller). This will, in turn, provide basic operations of the device. This
study could also be used in the supervision of a trained physiotherapist to provide functional
restoration to patients with spinal cord injury. In addition to that, this method can also be used in
sports Biomechanics.
REFERENCES
[1] S. R. Guntur, R. R. Gorrepati, and V. R. Dirisala, "Robotics in Healthcare : An Internet of Medical
Robotic Things ( IoMRT )" Perspective. Elsevier Inc., 2016.
[2] M. Radman and A. Chaibakhsh, “Generalized Sequential Forward Selection Method for Channel
Selection in EEG Signals for Classification of Left or Right Hand Movement in BCI,” ICCKE, pp.
137–142, 2019.
[3] N. E. Isa, A. Amir, M. Z. Ilyas, and M. S. Razalli, “Motor imagery classification in Brain- computer
interface ( BCI ) based on EEG signal by using machine learning technique”, Bulletin of Electrical
Engineering and Informatics, vol. 8, no. 1, pp. 269–275, 2019.
[4] E. Erkan and I. Kurnaz, “A study on the effect of psychophysiological signal features on
classification methods,” Measurement, vol. 101, pp. 45–52, 2017.
[5] J. Wang, “EEG signal classification with feature selection based on one-dimension real- valued
particle swarm optimization,” International Conference on Mechatronics, Control and Electronic
Engineering, pp. 310–314, 2014.
14. Signal & Image Processing: An International Journal (SIPIJ) Vol.12, No.6, December 2021
50
[6] K. Aswinseshadri and V. T. Bai, “Feature Selection In Brain-Computer Interface Using Genetics
Method,” 2015 IEEE Int. Conf. Comput. Inf. Technol. Ubiquitous Comput. Commun. Dependable,
Auton. Secur. Comput. Pervasive Intell. Comput., pp. 270–275, 2015.
[7] C. K. S. G. Ponnambalam, “Binary and multi-class motor imagery using Renyi entropy for feature
extraction,” Neural Comput. Appl., vol. 28, no. 8, pp. 2051–2062, 2017.
[8] K. Venkatachalam, A. Devipriya, J. Maniraj, M. Sivaram, A. Ambikapathy, and I. S. Amiri, “A novel
method of motor imagery classification using EEG signal”, Artificial Intelligence In Medicine, vol.
103, no. December 2019, 2020.
[9] R. Chatterjee, T. Maitra, S. K. Hafizul, and M. Mehedi, “A novel machine learning-based feature
selection for motor imagery EEG signal classification on the Internet of medical things environment,”
Futur. Gener. Comput. Syst., vol. 98, pp. 419–434, 2019.
[10] Liu, A., Chen, K., Liu, Q., Ai, Q., Xie, Y., Chen, A.: Feature Selection for Motor Imagery EEG
Classification Based on Firefly Algorithm and Learning Automata. Sensors. 17, 2576 (2017).
[11] M. Witkowski, M. Cortese, M. Cempini, J. Mellinger, and N. Vitiello, “Enhancing brain- machine
interface ( BMI ) control of a hand exoskeleton using electrooculography (EOG)”, Journal of Neuro
Engineering and Rehabilitation, pp. 1–6, 2014.
[12] J. R. Wolpaw, N. Birbaumer, D. J. McFarland, G. Pfurtscheller, and T. M. Vaughan, “Brain-
computer interfaces for communication and control,” Clin. Neurophysiol., vol. 113, no. 6, pp. 767–
791, 2002.
[13] D. J. McFarland, L. a Miner, T. M. Vaughan, and J. R. Wolpaw, “Mu and beta rhythm topographies
during motor imagery and actual movements.,” Brain Topogr., vol. 12, no. 3, pp. 177–186, 2000.
[14] G. Pfurtscheller, A. Stancák, and C. Neuper, “Event-related synchronization (ERS) in the alpha band
— an electrophysiological correlate of cortical idling: A review,” Int. J. Psychophysiol., vol. 24, no.
1, pp. 39–46, 1996.
[15] H. Bashashati, R. K. Ward, G. E. Birch, and A. Bashashati, “Comparing Different Classifiers in
Sensory Motor Brain-Computer Interfaces”, PLOS, pp. 1–17, 2015.
[16] Khan, Bilal Alam, Anam Hashmi, and Omar Farooq. "An Automated System for Epileptic Seizure
Detection Using EEG." In Advances in Data and Information Sciences, pp. 147-155. Springer,
Singapore, 2020.
[17] B. A. Khan, A. R. Usmani, S. Athar, A. Hashmi, O. Farooq, and M. Muzammil. "EEG-Based
Exoskeleton for Rehabilitation Therapy." In Ergonomics for Improved Productivity, pp. 645-653.
Springer, Singapore, 2021.
[18] P. K. Parashiva and A. P. Vinod, "A New Channel Selection Method using Autoencoder for Motor
Imagery based Brain Computer Interface," 2019 IEEE International Conference on Systems, Man and
Cybernetics (SMC), 2019, pp. 3641-3646, doi: 10.1109/SMC.2019.8914251.
[19] A. Hashmi, B. A. Khan, O. Farooq, “A study of the classification of Motor imagery signals using
Machine learning tools,” David C. Wyld et al. (Eds): VLSIA, INWES, CSIP, BMLI - 2021 pp. 23-33,
2021. CS & IT - CSCP 2021, doi: 10.5121/csit.2021.112104
[20] Pedregosa et al., “Scikit-learn: Machine Learning in Python,”Journal of Machine Learning Research,
vol. 12, pp. 2825-2830, 2011.