Common Spatial Pattern Filters Research Articles

Exploring EEG-based motor imagery decoding: a dual approach using spatial features and spectro-spatial Deep Learning model IFNet.

In recent years, the decoding of motor imagery (MI) from electroencephalography (EEG) signals has become a focus of research for brain-machine interfaces (BMIs) and neurorehabilitation. However, EEG signals present challenges due to their non-stationarity and the substantial presence of noise commonly found in recordings, making it difficult to design highly effective decoding algorithms. These algorithms are vital for controlling devices in neurorehabilitation tasks, as they activate the patient's motor cortex and contribute to their recovery. This study proposes a novel approach for decoding MI during pedalling tasks using EEG signals. A widespread approach is based on feature extraction using Common Spatial Patterns (CSP) followed by a linear discriminant analysis (LDA) as a classifier. The first approach covered in this work aims to investigate the efficacy of a task-discriminative feature extraction method based on CSP filter and LDA classifier. Additionally, the second alternative hypothesis explores the potential of a spectro-spatial Convolutional Neural Network (CNN) to further enhance the performance of the first approach. The proposed CNN architecture combines a preprocessing pipeline based on filter banks in the frequency domain with a convolutional neural network for spectro-temporal and spectro-spatial feature extraction. To evaluate the approaches and their advantages and disadvantages, EEG data has been recorded from several able-bodied users while pedalling in a cycle ergometer in order to train motor imagery decoding models. The results show levels of accuracy up to 80% in some cases. The CNN approach shows greater accuracy despite higher instability.

What are we really decoding? Unveiling biases in EEG-based decoding of the spatial focus of auditory attention

Objective. Spatial auditory attention decoding (Sp-AAD) refers to the task of identifying the direction of the speaker to which a person is attending in a multi-talker setting, based on the listener’s neural recordings, e.g. electroencephalography (EEG). The goal of this study is to thoroughly investigate potential biases when training such Sp-AAD decoders on EEG data, particularly eye-gaze biases and latent trial-dependent confounds, which may result in Sp-AAD models that decode eye-gaze or trial-specific fingerprints rather than spatial auditory attention. Approach. We designed a two-speaker audiovisual Sp-AAD protocol in which the spatial auditory and visual attention were enforced to be either congruent or incongruent, and we recorded EEG data from sixteen participants undergoing several trials recorded at distinct timepoints. We trained a simple linear model for Sp-AAD based on common spatial patterns filters in combination with either linear discriminant analysis (LDA) or k-means clustering, and evaluated them both across- and within-trial. Main results. We found that even a simple linear Sp-AAD model is susceptible to overfitting to confounding signal patterns such as eye-gaze and trial fingerprints (e.g. due to feature shifts across trials), resulting in artificially high decoding accuracies. Furthermore, we found that changes in the EEG signal statistics across trials deteriorate the trial generalization of the classifier, even when the latter is retrained on the test trial with an unsupervised algorithm. Significance. Collectively, our findings confirm that there exist subtle biases and confounds that can strongly interfere with the decoding of spatial auditory attention from EEG. It is expected that more complicated non-linear models based on deep neural networks, which are often used for Sp-AAD, are even more vulnerable to such biases. Future work should perform experiments and model evaluations that avoid and/or control for such biases in Sp-AAD tasks.

Multiband Convolutional Riemannian Network with Band-wise Riemannian Triplet Loss for Motor Imagery Classification.

This paper presents a novel motor imagery classification algorithm that uses an overlapping multiscale multiband convolutional Riemannian network with band-wise Riemannian triplet loss to improve classification performance. Despite the superior performance of the Riemannian approach over the common spatial pattern filter approach, deep learning methods that generalize the Riemannian approach have received less attention. The proposed algorithm develops a state-of-the-art multiband Riemannian network that reduces the potential overfitting problem of Riemannian networks, a drawback of Riemannian networks due to their inherent large feature dimension from covariance matrix, by using fewer subbands with discriminative frequency diversity, by inserting convolutional layers before computing the subband covariance matrix, and by regularizing subband networks with Riemannian triplet loss. The proposed method is evaluated using the publicly available datasets, the BCI Competition IV dataset 2a and the OpenBMI dataset. The experimental results confirm that the proposed method improves performance, in particular achieving state-of-the-art classification accuracy among the currently studied Riemannian networks.

Enhancing motor imagery detection efficacy using multisensory virtual reality priming.

Brain-computer interfaces (BCI) have been developed to allow users to communicate with the external world by translating brain activity into control signals. Motor imagery (MI) has been a popular paradigm in BCI control where the user imagines movements of e.g., their left and right limbs and classifiers are then trained to detect such intent directly from electroencephalography (EEG) signals. For some users, however, it is difficult to elicit patterns in the EEG signal that can be detected with existing features and classifiers. As such, new user control strategies and training paradigms have been highly sought-after to help improve motor imagery performance. Virtual reality (VR) has emerged as one potential tool where improvements in user engagement and level of immersion have shown to improve BCI accuracy. Motor priming in VR, in turn, has shown to further enhance BCI accuracy. In this pilot study, we take the first steps to explore if multisensory VR motor priming, where haptic and olfactory stimuli are present, can improve motor imagery detection efficacy in terms of both improved accuracy and faster detection. Experiments with 10 participants equipped with a biosensor-embedded VR headset, an off-the-shelf scent diffusion device, and a haptic glove with force feedback showed that significant improvements in motor imagery detection could be achieved. Increased activity in the six common spatial pattern filters used were also observed and peak accuracy could be achieved with analysis windows that were 2 s shorter. Combined, the results suggest that multisensory motor priming prior to motor imagery could improve detection efficacy.

EEG-based BCI: A novel improvement for EEG signals classification based on real-time preprocessing

This work aims to improve EEG signal binary and multiclass classification for real-time BCI applications. Therefore, our paper discusses the results of a new real-time approach that was integrated into a complete prediction system, where we proposed a new trick to eliminate the effect of EEG’s non-stationarity nature. This improvement can increase the accuracy from 50% using raw EEG to the order of 90% after preprocessing step in the binary case and from 28% to 78% in the multiclass case. Then, we chose to filter all signals by the proposed bandpass filter automatically optimized using the sine cosine algorithm (SCA) to find the optimal bandwidth that contains the entire EEG characteristics in beta waves. Moreover, we used a common spatial pattern (CSP) filter to eliminate the correlation between all extracted features. Then, the light gradient boosting machine (LGBM) classifier is also combined with SCA algorithm to build better prediction models. As a result, the outcome system was applied on UCI and PhysioNet datasets to get excellent accuracy values of higher than 99% and 95%, respectively, using the data acquired only from three channels. On the other hand, the related works used all the data acquired from 14 channels to find an accuracy value between 70% and 98.5%, which shows the robustness of our method to improve EEG signal prediction quality.

An Intelligent Fault Diagnosis Method of Rolling Bearings via Variational Mode Decomposition and Common Spatial Pattern-Based Feature Extraction

Monitoring and identifying the health condition of rolling bearings can reduce the risk of mechanical equipment failure. This paper proposes a novel intelligent diagnosis method of rolling bearings: First, the vibration signals are decomposed into band-limited instinct mode functions (BLIMFs) by variational mode decomposition (VMD). Then, the proposed high-dimensional common spatial pattern (hdCSP) filter is used to generate the high-dimensional eigenvectors representing the decomposed BLIMFs. Finally, the random forests classifier is used to classify the eigenvectors and obtain the diagnosis results. The performance of the proposed VMD-hdCSP method is evaluated on the Case Western Reserve University dataset. The experimental results show the proposed method can automatically classify different health states of rolling bearings and obtain precise diagnosis results.

Cross-Platform Implementation of an SSVEP-Based BCI for the Control of a 6-DOF Robotic Arm.

Robotics has been successfully applied in the design of collaborative robots for assistance to people with motor disabilities. However, man-machine interaction is difficult for those who suffer severe motor disabilities. The aim of this study was to test the feasibility of a low-cost robotic arm control system with an EEG-based brain-computer interface (BCI). The BCI system relays on the Steady State Visually Evoked Potentials (SSVEP) paradigm. A cross-platform application was obtained in C++. This C++ platform, together with the open-source software Openvibe was used to control a Stäubli robot arm model TX60. Communication between Openvibe and the robot was carried out through the Virtual Reality Peripheral Network (VRPN) protocol. EEG signals were acquired with the 8-channel Enobio amplifier from Neuroelectrics. For the processing of the EEG signals, Common Spatial Pattern (CSP) filters and a Linear Discriminant Analysis classifier (LDA) were used. Five healthy subjects tried the BCI. This work allowed the communication and integration of a well-known BCI development platform such as Openvibe with the specific control software of a robot arm such as Stäubli TX60 using the VRPN protocol. It can be concluded from this study that it is possible to control the robotic arm with an SSVEP-based BCI with a reduced number of dry electrodes to facilitate the use of the system.

A neurophysiological approach to spatial filter selection for adaptive brain–computer interfaces

Objective. The common spatial patterns (CSP) algorithm is an effective method to extract discriminatory features from electroencephalography (EEG) to be used by a brain–computer interface (BCI). However, informed selection of CSP filters typically requires oversight from a BCI expert to accept or reject filters based on the neurophysiological plausibility of their activation patterns. Our goal was to identify, analyze and automatically classify prototypical CSP patterns to enhance the prediction of motor imagery states in a BCI. Approach. A data-driven approach that used four publicly available EEG datasets was adopted. Cluster analysis revealed recurring, visually similar CSP patterns and a convolutional neural network was developed to distinguish between established CSP pattern classes. Furthermore, adaptive spatial filtering schemes that utilize the categorization of CSP patterns were proposed and evaluated. Main results. Classes of common neurophysiologically probable and improbable CSP patterns were established. Analysis of the relationship between these categories of CSP patterns and classification performance revealed discarding neurophysiologically improbable filters can decrease decoder performance. Further analysis revealed that the spatial orientation of EEG modulations can evolve over time, and that the features extracted from the original CSP filters can become inseparable. Importantly, it was shown through a novel adaptive CSP technique that adaptation in response to these emerging patterns can restore feature separability. Significance. These findings highlight the importance of considering and reporting on spatial filter activation patterns in both online and offline studies. They also emphasize to researchers in the field the importance of spatial filter adaptation in BCI decoder design, particularly for online studies with a focus on training users to develop stable and suitable brain patterns.

An improved version of local activities estimation to enhance motor imagery classification

ObjectiveThe common spatial pattern (CSP) and its variants are popular in the EEG-based motor imagery BCIs. However, this method has some drawbacks, especially when a few labeled samples are available. The local activities estimation (LAE) method works well with small training sets, but it is more sensitive to the position of the electrodes. Here, we suggest a combination of the LAE and CSP, namely LAE-CSP, which performs better than both methods. MethodsIn LAE-CSP, the EEG signal passes through regularized CSP and LAE spatial filters after band-pass filtering and then the data dimension are reduced based on the physiological data. Afterwards, the features are extracted using fast Fourier transform (FFT). In this work, CSP, FBRCSP, EA-CSP, LAE and LAE-CSP, methods were evaluated and compared. Three sets of motor imagery data from BCI competition III and IV along with Cho et al. dataset, including EEG signals from 28 subjects were used in this study. For each dataset, the training set were selected in 21 different sizes. ResultsLAE-CSP performs better than all tested methods. Particularly it has good performance using only ten labeled samples per class, with an average accuracy of almost 80 %. ConclusionsLAE-CSP reliably enhances the performance of the motor imagery-based brain-computer interfaces. SignificanceThe LAE-CSP takes advantage of the LAE and CSP and compensates for the drawbacks of both methods.

Surface electromyography and electroencephalography processing in dysarthric patients for verbal commands or speaking intention characterization

Assistance systems for people with Cerebrovascular Accident (CVA) after-effects such as dysarthria are gaining interest due to the increasing proportion of the population with these disorders. This paper proposes the acquisition and processing of surface Electromyography (sEMG) on the facial muscles activated during the process of speaking and Electroencephalography (EEG), synchronized in time by means of an audio file. Data was collected in seven healthy volunteers and seven patients with dysarthria at the Physiotherapy Department of the Porto Alegre Clinical Hospital. The main objective is to classify the biosignals for verbal commands established by Support Vector Machine (SVM) method for the sEMG and Naïve Bayes (NB) method for the EEG. Both methods were compared with the Linear Discriminant Analysis (LDA) method. Extracted features of sEMG signal were: standard deviation, arithmetic mean, skewness, kurtosis and RMS. For the EEG signal, characteristics were extracted in frequency domain such as: minimum, maximum, average and standard deviation, and skewness and kurtosis, in the time domain. As part of the processing method, Common Spatial Pattern (CSP) filters were employed in order to increase the discriminating activity between motion classes in the EEG signal. Mental task identification in healthy and dysarthric subjects reached an accuracy of 77.6–80.8%.

Sequence-to-sequence deep neural network with spatio-spectro and temporal features for motor imagery classification

Electroencephalography (EEG) is a method of the brain–computer interface (BCI) that measures brain activities. EEG is a method of (non-)invasive recording of the electrical activity of the brain. This can be used to build BCIs. From the last decade, EEG has grasped researchers’ attention to distinguish human activities. However, temporal information has rarely been retained to incorporate temporal information for multi-class (more than two classes) motor imagery classification. This research proposes a long-short-term-memory-based deep learning model to learn the hidden sequential patterns. Two types of features are used to feed the proposed model, including Fourier Transform Energy Maps (FTEMs) and Common Spatial Patterns (CSPs) filters. Multiple experiments have been conducted on a publicly available dataset. Extraction of spatial and spectro-temporal features using CSP filters and FTEM allow the sequence-to-sequence based proposed model to learn the hidden sequential features. The proposed method is trained, evaluated, and optimized for a publicly available benchmark data set and resulted in 0.81 mean kappa value. Obtained results depict the model robustness for the artifacts and suitable for real-life applications with comparable classification accuracy. The code and findings will be available at https://github.com/waseemabbaas/Motor-Imagery-Classification.git.

Fast EEG-Based Decoding Of The Directional Focus Of Auditory Attention Using Common Spatial Patterns.

Noise reduction algorithms in current hearing devices lack informationabout the sound source a user attends to when multiple sources are present. To resolve this issue, they can be complemented with auditory attention decoding (AAD) algorithms, which decode the attention using electroencephalography (EEG) sensors. State-of-the-art AAD algorithms employ a stimulus reconstruction approach, in which the envelope of the attended source is reconstructed from the EEG and correlated with the envelopes of the individual sources. This approach, however, performs poorly on short signal segments, whilelonger segments yield impractically long detection delays when the user switches attention. We propose decoding the directional focus of attention using filterbank common spatial pattern filters (FB-CSP) as an alternative AAD paradigm, whichdoes not require access to the clean source envelopes. The proposed FB-CSP approach outperforms both the stimulus reconstruction approach on short signal segments, as well as a convolutional neural network approach on the same task. We achieve a high accuracy (80% for [Formula: see text] windows and 70% for quasi-instantaneous decisions), which is sufficient to reach minimal expected switch durations below [Formula: see text]. We also demonstrate that the decoder can adapt to unlabeled data from anunseen subject and works with only a subset of EEG channels located around the ear to emulate a wearable EEG setup. The proposed FB-CSP method provides fast and accurate decoding of the directional focus of auditory attention. The high accuracy on very short data segments is a major step forward towards practical neuro-steered hearing devices.

Vocal Imagery vs Intention: Viability of Vocal-Based EEG-BCI Paradigms.

The viability of electroencephalogram (EEG) based vocal imagery (VIm) and vocal intention (VInt) Brain-Computer Interface (BCI) systems has been investigated in this study. Four different types of experimental tasks related to humming has been designed and exploited here. They are: (i) non-task specific (NTS), (ii) motor task (MT), (iii) VIm task, and (iv) VInt task. EEG signals from seventeen participants for each of these tasks were recorded from 16 electrode locations on the scalp and its features were extracted and analysed using common spatial pattern (CSP) filter. These features were subsequently fed into a support vector machine (SVM) classifier for classification. This analysis aimed to perform a binary classification, predicting whether the subject was performing one task or the other. Results from an extensive analysis showed a mean classification accuracy of 88.9% for VIm task and 91.1% for VInt task. This study clearly shows that VIm can be classified with ease and is a viable paradigm to integrate in BCIs. Such systems are not only useful for people with speech problems, but in general for people who use BCI systems to help them out in their everyday life, giving them another dimension of system control.

Multiclass Classification of Spatially Filtered Motor Imagery EEG Signals Using Convolutional Neural Network for BCI Based Applications

Brain–Computer Interface (BCI) system offers a new means of communication for those with paralysis or severe neuromuscular disorders. BCI systems based on Motor Imagery (MI) Electroencephalography (EEG) signals enable the user to convert their thoughts into actions without any voluntary muscle movement. Recently, Convolutional neural network (CNN) is used for the classification of MI signals. However, to produce good MI classification, it is necessary to effectively represent the signal as an input image to the CNN and train the deep learning classifier using large training data. In this work, EEG signals are acquired over 16 channels and are filtered using a bandpass filter with the frequency range of 1 to 100 Hz. The processed signal is spatially filtered using Common Spatial Pattern (CSP) filter. The spectrograms of the spatially filtered signals are given as input to CNN. A single convolutional layer CNN is designed to classify left hand, right hand, both hands, and feet MI EEG signals. The size of the training data is increased by augmenting the spectrograms of the EEG signals. The CNN classifier was evaluated using MI signals acquired from twelve healthy subjects. Results show that the proposed method achieved an average classification accuracy of 95.18 ± 2.51% for two-class (left hand and right hand) and 87.37 ± 1.68% for four-class (Left hand, Right hand, Both hands, and Feet) MI. Thus, the method manifests that this 2D representation of 1D EEG signal along with image augmentation shows a high potential for classification of MI EEG signals using the designed CNN model.

Brain Computer Interface Classifiers for Semi-Autonomous Wheelchair Using Fuzzy Logic Optimization

Abstract A brain–computer interface (BCI)-based controller bridges the gap between smart wheelchairs and physically impaired persons with severe conditions. This paper presents the design of a hybrid BCI controller with six classifiers using an electroencephalogram (EEG) headset to detect hand motor imagery (MI) and jaw electromyography (EMG) signals. A BCI controller and semi-autonomous system is developed to control a smart wheelchair in conjunction with its semi-autonomous capabilities. For data acquisition, an openvibe system and a commercial grade EEG headset are used. A multiple common spatial pattern (CSP) filter and Linear discriminant analysis (LDA) classifier system is used to process and classify the user's brain activity. To convert the classifier data into a signal that is compatible with the semi-autonomous wheelchair system, a fuzzy logic controller (FLC) is integrated in LabVIEW. Subjects are trained to use the BCI system and the classifier profiles are optimized for each user and the results are analyzed for this study. The openvibe “Replay” script and recorded training data are used to evaluate the performance of the controller scheme. For each subject, positive, negative, and false-positive executions are recorded. During the initial testing phase, the positive rates for subjects were strong, but false-positive rates were too high to be used. Therefore, the design is iterated by changing the rules of the FLC and configuration of the LabVIEW script. The configuration with the best positive rates for turn executions is chosen where the average positive rate for turning is 0.68 for subject 1 and 0.64 for subject 2.

Optimal Selection of EEG Electrodes Using Interval Type-2 Fuzzy-Logic-Based Semiseparating Signaling Game.

This article addresses the noise contamination in spatial filtering of brain responses using a novel signaling game-based approach to the optimal selection of EEG electrodes. The proposed method takes the standard common spatial pattern (CSP) filter as an input and produces an optimal electrode set as output for effective classification of different cognitive tasks. The standard CSP algorithms are highly prone to the inclusion of noise in the EEG data and may select noisy electrodes/signal sources that are redundant for a specific cognitive task which, in turn, may lead to a lower classification accuracy. A lot of literature exists in this area of research, most of which deals with adding the regularization term in the standard CSP algorithm. However, all of these methods lack capturing the uncertainty present in the EEG responses due to intrasession and intersession variations of subjective brain response. The novelty of this article lies in designing the fuzzy signaling game-based approach for optimal electrode selection using an interval type-2 fuzzy set, which can capture both the intrasession and intersession variability of EEG responses acquired from a subject's scalp. Experiments are undertaken over a wide variety of possible cognitive task classification problems which reveal that the proposed method yields superior results in electrode selection with respect to classification accuracy. Statistical tests undertaken using the Friedman test also confirm the superiority of the proposed method over its competitors.

An Adaptive CSP and Clustering Classification for Online Motor Imagery EEG

A potential limitation of motor imagery (MI) based brain-computer interface (BCI) (MI-BCI) is that it usually requires a relatively long time to record sufficient electroencephalogram (EEG) data for robust feature extraction and classification. Moreover, due to the non-stationarities in EEG signals, the offline training model has poor adaptability and classification ability in cross-session or sample-wise online testing. Methods: To address the problems, we propose a model updating scheme with adaptive and fast operation. Based on the Common Spatial Pattern (CSP), we propose an online and fast generalized eigendecomposition method by Recursive Least Squares updates of the CSP filter coefficients (RLS-CSP), which allows incremental training for CSP spatial filters. Additionally, we present an Incremental Self-training Classification algorithm based on Density Clustering (ISCDC) to select high-confidence samples to update spatial filters and classifier, and classify at the same time. Results: We conducted extensive experiments to validate the efficiency of the proposed adaptive CSP and classifier on the BCI III_IVa and BCI III_V data sets. Experimental results demonstrate that RLS-CSP outperforms significantly in a small sample setting (SSS), and ISCDC has great adaptability in cross-session and non-stationary EEG signals. The results indicate that our proposed methods are feasible to improve the real-time performance of online BCI system.

Classification of left and right foot kinaesthetic motor imagery using common spatial pattern

Background and objectives: Brain-computer interface (BCI) systems typically deploy common spatial pattern (CSP) for feature extraction of mu and beta rhythms based on upper-limbs kinaesthetic motor imageries (KMI). However, it was not used to classify the left versus right foot KMI, due to its location inside the mesial wall of sensorimotor cortex, which makes it difficult to be detected. We report novel classification of mu and beta EEG features, during left and right foot KMI cognitive task, using CSP, and filter bank common spatial pattern (FBCSP) method, to optimize the subject-specific band selection. We initially proposed CSP method, followed by the implementation of FBCSP for optimization of individual spatial patterns, wherein a set of CSP filters was learned, for each of the time/frequency filters in a supervised way. This was followed by the log-variance feature extraction and concatenation of all features (over all chosen spectral-filters). Subsequently, supervised machine learning was implemented, i.e. logistic regression (Logreg) and linear discriminant analysis (LDA), in order to compare the respective foot KMI classification rates. Training and testing data, used in the model, was validated using 10-fold cross validation. Four methodology paradigms are reported, i.e. CSP LDA, CSP Logreg, and FBCSP LDA, FBCSP Logreg. All paradigms resulted in an average classification accuracy rate above the statistical chance level of 60.0% (P < 0.01). On average, FBCSP LDA outperformed remaining paradigms with kappa score of 0.41 and classification accuracy of 70.28% ± 4.23. Similarly, this paradigm enabled discrimination between right and left foot KMI cognitive task at highest accuracy rate i.e. maximum 77.5% with kappa = 0.55 and the area under ROC curve as 0.70 (in single-trial analysis). The proposed novel paradigms, using CSP and FBCSP, established a potential to exploit the left versus right foot imagery classification, in synchronous 2-class BCI for controlling robotic foot, or foot neuroprosthesis.

Motion Recognition and an Accuracy Comparison of Left and Right Arms by EEG Signal Analysis

An electromyogram (EMG) is a signal for muscle output that indicates the degree of muscle contraction and relaxation. For these muscle signals to be output, certain signals must be received from the brain. To analyze these relations, electroencephalograms (EEGs) of the brain are measured to extract brain waves that are active at that time, although it is difficult to identify or distinguish expression patterns of the brain signal through EMG output. However, the brain signal operates via a partially reached signal and transmits the results of the operation. In this study, we analyze signals transmitted in this process and confirm whether human motion can be predicted from brain signals. It is not easy to guess the exact protocol of the brain using a general method, because a biosignal is a signal that differs from person to person. However, by analyzing the signals displayed by a particular user through actions, it is possible to determine the presence or absence of a signal to distinguish muscle movements. In the course of signal transduction, the energy of the left and right brain waves changes in the form of energy or signals that cause an arm’s movement. Responding to this, we analyze the signal transmission process of brain signals and EMGs to analyze loss and generated output. We extract EEG data from brain waves and determine EMG signals from the energy characteristics; we then collect and merge the results of spectra analysis through the Common Spatial Pattern (CSP) filter and explore the basis for predicting wills during muscle signals and stimulation transmission. The active information of the data within the working time of left and right brain waves depends on the changes of the left and right brain waves. It is proposed that the appearance of similar signals at these specific timescales can help identify the operations of the arms and outputs by the left and right biceps.

Classification of Motor Functions from Electroencephalogram (EEG) Signals Based on an Integrated Method Comprised of Common Spatial Pattern and Wavelet Transform Framework.

In this work, an algorithm for the classification of six motor functions from an electroencephalogram (EEG) signal that combines a common spatial pattern (CSP) filter and a continuous wavelet transform (CWT), is investigated. The EEG data comprise six grasp-and-lift events, which are used to investigate the potential of using EEG as input signals with brain computer interface devices for controlling prosthetic devices for upper limb movement. Selected EEG channels are the ones located over the motor cortex, C3, Cz and C4, as well as at the parietal region, P3, Pz and P4. In general, the proposed algorithm includes three main stages, band pass filtering, CSP filtering, and wavelet transform and training on GoogLeNet for feature extraction, feature learning and classification. The band pass filtering is performed to select the EEG signal in the band of 7 Hz to 30 Hz while eliminating artifacts related to eye blink, heartbeat and muscle movement. The CSP filtering is applied on two-class EEG signals that will result in maximizing the power difference between the two-class dataset. Since CSP is mathematically developed for two-class events, the extension to the multiclass paradigm is achieved by using the approach of one class versus all other classes. Subsequently, continuous wavelet transform is used to convert the band pass and CSP filtered signals from selected electrodes to scalograms which are then converted to images in grayscale format. The three scalograms from the motor cortex regions and the parietal region are then combined to form two sets of RGB images. Next, these RGB images become the input to GoogLeNet for classification of the motor EEG signals. The performance of the proposed classification algorithm is evaluated in terms of precision, sensitivity, specificity, accuracy with average values of 94.8%, 93.5%, 94.7%, 94.1%, respectively, and average area under the receiver operating characteristic (ROC) curve equal to 0.985. These results indicate a good performance of the proposed algorithm in classifying grasp-and-lift events from EEG signals.