Upper limb complex movements decoding from pre-movement EEG signals using wavelet common spatial patterns

EEG-based motor imagery channel selection and classification using hybrid optimization and two-tier deep learning

Brain-computer interface (BCI) technology holds promise for individuals with profound motor impairments, offering the potential for communication and control. Motor imagery (MI)-based BCI systems are particularly relevant in this context. Despite their potential, achieving accurate and robust classification of MI tasks using electroencephalography (EEG) data remains a significant challenge. In this paper, we employed the Minimum Redundancy Maximum Relevance (MRMR) algorithm to optimize channel selection. Furthermore, we introduced a hybrid optimization approach that combines the War Strategy Optimization (WSO) and Chimp Optimization Algorithm (ChOA). This hybridization significantly enhances the classification model's overall performance and adaptability. A two-tier deep learning architecture is proposed for classification, consisting of a Convolutional Neural Network (CNN) and a modified Deep Neural Network (M-DNN). The CNN focuses on capturing temporal correlations within EEG data, while the M-DNN is designed to extract high-level spatial characteristics from selected EEG channels. Integrating optimal channel selection, hybrid optimization, and the two-tier deep learning methodology in our BCI framework presents an enhanced approach for precise and effective BCI control. Our model got 95.06% accuracy with high precision. This advancement has the potential to significantly impact neurorehabilitation and assistive technology applications, facilitating improved communication and control for individuals with motor impairments.

A brain machine interface framework for exploring proactive control of smart environments

Brain machine interfaces (BMIs) can substantially improve the quality of life of elderly or disabled people. However, performing complex action sequences with a BMI system is onerous because it requires issuing commands sequentially. Fundamentally different from this, we have designed a BMI system that reads out mental planning activity and issues commands in a proactive manner. To demonstrate this, we recorded brain activity from freely-moving monkeys performing an instructed task and decoded it with an energy-efficient, small and mobile field-programmable gate array hardware decoder triggering real-time action execution on smart devices. Core of this is an adaptive decoding algorithm that can compensate for the day-by-day neuronal signal fluctuations with minimal re-calibration effort. We show that open-loop planning-ahead control is possible using signals from primary and pre-motor areas leading to significant time-gain in the execution of action sequences. This novel approach provides, thus, a stepping stone towards improved and more humane control of different smart environments with mobile brain machine interfaces.

Evaluation of temporal, spatial and spectral filtering in CSP-based methods for decoding pedaling-based motor tasks using EEG signals

Stroke is a neurological syndrome that usually causes a loss of voluntary control of lower/upper body movements, making it difficult for affected individuals to perform Activities of Daily Living (ADLs). Brain-Computer Interfaces (BCIs) combined with robotic systems, such as Motorized Mini Exercise Bikes (MMEB), have enabled the rehabilitation of people with disabilities by decoding their actions and executing a motor task. However, Electroencephalography (EEG)-based BCIs are affected by the presence of physiological and non-physiological artifacts. Thus, movement discrimination using EEG become challenging, even in pedaling tasks, which have not been well explored in the literature. In this study, Common Spatial Patterns (CSP)-based methods were proposed to classify pedaling motor tasks. To address this, Filter Bank Common Spatial Patterns (FBCSP) and Filter Bank Common Spatial-Spectral Patterns (FBCSSP) were implemented with different spatial filtering configurations by varying the time segment with different filter bank combinations for the three methods to decode pedaling tasks. An in-house EEG dataset during pedaling tasks was registered for 8 participants. As results, the best configuration corresponds to a filter bank with two filters (8-19 Hz and 19-30 Hz) using a time window between 1.5 and 2.5 s after the cue and implementing two spatial filters, which provide accuracy of approximately 0.81, False Positive Rates lower than 0.19, andKappaindex of 0.61. This work implies that EEG oscillatory patterns during pedaling can be accurately classified using machine learning. Therefore, our method can be applied in the rehabilitation context, such as MMEB-based BCIs, in the future.

Effect of the period of EEG signals on the decoding of motor information

Decoding movement information from electroencephalogram to construct brain-computer interface has promising applications. The EEG data during the entire motor imagery (MI) period or movement execution (ME) period is generally decoded, and calculation of numerous information and massive dataset is time-consuming. In order to improve decoding efficiency, the joint topographic maps of the brain activation state of 15 subjects were studied during different periods. The results showed that the activation intensity of the preparation period in the motor imagery experiment was higher than during the exercise period, while during the exercise period, the activation intensity was higher than in the preparation period in the movement execution experiment. Hence, the wavelet neural network was used to decode the six-class movements including elbow flexion/extension, forearm pronation/supination and hand open/close in periods of MI/ME. The experimental results show that the accuracy obtained in the preparation period is the highest in the motor imagery experiment, which is 80.77%. On the other hand, the highest accuracy obtained in the exercise period of the movement execution experiment is 79.26%. It further proves that the optimized period is a key decoding factor to reduce the cost of calculation, and this new decoding method is effective to build a more intelligent brain-computer interface system.

A Combined Virtual Electrode-Based ESA and CNN Method for MI-EEG Signal Feature Extraction and Classification

A Brain-Computer Interface (BCI) is a medium for communication between the human brain and computers, which does not rely on other human neural tissues, but only decodes Electroencephalography (EEG) signals and converts them into commands to control external devices. Motor Imagery (MI) is an important BCI paradigm that generates a spontaneous EEG signal without external stimulation by imagining limb movements to strengthen the brain's compensatory function, and it has a promising future in the field of computer-aided diagnosis and rehabilitation technology for brain diseases. However, there are a series of technical difficulties in the research of motor imagery-based brain-computer interface (MI-BCI) systems, such as: large individual differences in subjects and poor performance of the cross-subject classification model; a low signal-to-noise ratio of EEG signals and poor classification accuracy; and the poor online performance of the MI-BCI system. To address the above problems, this paper proposed a combined virtual electrode-based EEG Source Analysis (ESA) and Convolutional Neural Network (CNN) method for MI-EEG signal feature extraction and classification. The outcomes reveal that the online MI-BCI system developed based on this method can improve the decoding ability of multi-task MI-EEG after training, it can learn generalized features from multiple subjects in cross-subject experiments and has some adaptability to the individual differences of new subjects, and it can decode the EEG intent online and realize the brain control function of the intelligent cart, which provides a new idea for the research of an online MI-BCI system.

Hybrid Deep Learning Approach for Stress Detection Using Decomposed EEG Signals

Stress has an impact, not only on a person's physical health, but also on the ability to perform at the workplace in daily life. The well-established relation between psychological stress and its pathogeneses highlights the need for detecting psychological stress early, in order to prevent disease advancement and to save human lives. Electroencephalography (EEG) signal recording tools are widely used to collect these psychological signals/brain rhythms in the form of electric waves. The aim of the current research was to apply automatic feature extraction to decomposed multichannel EEG recordings, in order to efficiently detect psychological stress. The traditional deep learning techniques, namely the convolution neural network (CNN), long short-term memory (LSTM), bidirectional long short-term memory (BiLSTM), gated recurrent unit (GRU) and recurrent neural network (RNN) models, have been frequently used for stress detection. A hybrid combination of these techniques may provide improved performance, and can handle long-term dependencies in non-linear brain signals. Therefore, this study proposed an integration of deep learning models, called DWT-based CNN, BiLSTM, and two layers of a GRU network, to extract features and classify stress levels. Discrete wavelet transform (DWT) analysis was used to remove the non-linearity and non-stationarity from multi-channel (14 channel) EEG recordings, and to decompose them into different frequency bands. The decomposed signals were utilized for automatic feature extraction using the CNN, and the stress levels were classified using BiLSTM and two layers of GRU. This study compared five combinations of the CNN, LSTM, BiLSTM, GRU and RNN models with the proposed model. The proposed hybrid model performed better in classification accuracy compared to the other models. Therefore, hybrid combinations are appropriate for the clinical intervention and prevention of mental and physical problems.

Decoding locomotion speed and slope from local field potentials of rat motor cortex

BACKGROUND AND OBJECTIVE

Local Field Potentials (LFPs) recorded from the primary motor cortex (M1) have been shown to be very informative for decoding movement parameters, and these signals can be used to decode forelimb kinematic and kinetic parameters accurately. Although locomotion is one of the most basic and important motor abilities of humans and animals, the potential of LFPs in decoding abstract hindlimb locomotor parameters has not been investigated. This study investigates the feasibility of decoding speed and slope of locomotion, as two important abstract parameters of walking, using the LFP signals.

METHODS

Rats were trained to walk smoothly on a treadmill with different speeds and slopes. The brain signals were recorded using the microwire arrays chronically implanted in the hindlimb area of M1 while rats walked on the treadmill. LFP channels were spatially filtered using optimal common spatial patterns to increase the discriminability of speeds and slopes of locomotion. Logarithmic wavelet band powers were extracted as basic features, and the best features were selected using the statistical dependency criterion before classification.

RESULTS

Using 5 s LFP trials, the average classification accuracies of four different speeds and seven different slopes reached 90.8% and 86.82%, respectively. The high-frequency LFP band (250-500 Hz) was the most informative band about these parameters and contributed more than other frequency bands in the final decoder model.

CONCLUSIONS

Our results show that the LFP signals in M1 accurately decode locomotion speed and slope, which can be considered as abstract walking parameters needed for designing long-term brain-computer interfaces for hindlimb locomotion control.

Metric Learning in Freewill EEG Pre-Movement and Movement Intention Classification for Brain Machine Interfaces

Decoding movement related intentions is a key step to implement BMIs. Decoding EEG has been challenging due to its low spatial resolution and signal to noise ratio. Metric learning allows finding a representation of data in a way that captures a desired notion of similarity between data points. In this study, we investigate how metric learning can help finding a representation of the data to efficiently classify EEG movement and pre-movement intentions. We evaluate the effectiveness of the obtained representation by comparing classification the performance of a Support Vector Machine (SVM) as a classifier when trained on the original representation, called Euclidean, and representations obtained with three different metric learning algorithms, including Conditional Entropy Metric Learning (CEML), Neighborhood Component Analysis (NCA), and the Entropy Gap Metric Learning (EGML) algorithms. We examine different types of features, such as time and frequency components, which input to the metric learning algorithm, and both linear and non-linear SVM are applied to compare the classification accuracies on a publicly available EEG data set for two subjects (Subject B and C). Although metric learning algorithms do not increase the classification accuracies, their interpretability using an importance measure we define here, helps understanding data organization and how much each EEG channel contributes to the classification. In addition, among the metric learning algorithms we investigated, EGML shows the most robust performance due to its ability to compensate for differences in scale and correlations among variables. Furthermore, from the observed variations of the importance maps on the scalp and the classification accuracy, selecting an appropriate feature such as clipping the frequency range has a significant effect on the outcome of metric learning and subsequent classification. In our case, reducing the range of the frequency components to 0–5 Hz shows the best interpretability in both Subject B and C and classification accuracy for Subject C. Our experiments support potential benefits of using metric learning algorithms by providing visual explanation of the data projections that explain the inter class separations, using importance. This visualizes the contribution of features that can be related to brain function.

Assessing residual motor function in patients with disorders of consciousness by brain network properties of task-state EEG

Recent achievements in evaluating the residual consciousness of patients with disorders of consciousness (DOCs) have demonstrated that spontaneous or evoked electroencephalography (EEG) could be used to improve consciousness state diagnostic classification. Recent studies showed that the EEG signal of the task-state could better characterize the conscious state and cognitive ability of the brain, but it has rarely been used in consciousness assessment. A cue-guide motor task experiment was designed, and task-state EEG were collected from 18 patients with unresponsive wakefulness syndrome (UWS), 29 patients in a minimally conscious state (MCS), and 19 healthy controls. To obtain the markers of residual motor function in patients with DOC, the event-related potential (ERP), scalp topography, and time-frequency maps were analyzed. Then the coherence (COH) and debiased weighted phase lag index (dwPLI) networks in the delta, theta, alpha, beta, and gamma bands were constructed, and the correlations of network properties and JFK Coma Recovery Scale-Revised (CRS-R) motor function scores were calculated. The results showed that there was an obvious readiness potential (RP) at the Cz position during the motor preparation process in the MCS group, but no RP was observed in the UWS group. Moreover, the node degree properties of the COH network in the theta and alpha bands and the global efficiency properties of the dwPLI network in the theta band were significantly greater in the MCS group compared to the UWS group. The above network properties and CRS-R motor function scores showed a strong linear correlation. These findings demonstrated that the brain network properties of task-state EEG could be markers of residual motor function of DOC patients.

Supplementary Information

The online version contains supplementary material available at 10.1007/s11571-021-09741-7.

Dynamic corticomuscular multi-regional modulations during finger movement revealed by time-varying network analysis

OBJECTIVE

A body movement involves the complicated information exchange between the central and peripheral systems, which is characterized by the dynamical coupling patterns between the multiple brain areas and multiple muscle units. How the central and peripheral nerves coordinate multiple internal brain regions and muscle groups is very important when accomplishing the action.

APPROACH

In this study, we extend the adaptive directed transfer function to construct the time-varying networks between multiple corticomuscular regions and divide the movement duration into different stages by the time-varying corticomuscular network patterns.

MAIN RESULTS

The inter dynamical corticomuscular network demonstrated the different interaction patterns between the central and peripheral systems during the different hand movement stages. The muscles transmit bottom-up movement information in the preparation stage, but the brain issues top-down control commands and dominates in the execution stage, and finally, the brain's dominant advantage gradually weakens in the relaxation stage. When classifying the different movement stages based on time-varying corticomuscular network indicators, an average accuracy above 74% could be reliably achieved.

SIGNIFICANCE

The findings of this study help deepen our knowledge of central-peripheral nerve pathways and coordination mechanisms, and also provide opportunities for monitoring and regulating movement disorders.

Feature- and classification analysis for detection and classification of tongue movements from single-trial pre-movement EEG

Individuals with severe tetraplegia can benefit from brain-computer interfaces (BCIs). While most movement-related BCI systems focus on right/left hand and/or foot movements, very few studies have considered tongue movements to construct a multiclass BCI. The aim of this study was to decode four movement directions of the tongue (left, right, up, and down) from single-trial pre-movement EEG and provide a feature and classifier investigation. In offline analyses (from ten healthy participants) detection and classification were performed using temporal, spectral, entropy, and template features classified using either a linear discriminative analysis, support vector machine, random forest or multilayer perceptron classifiers. Besides the 4-class classification scenario, all possible 3-, and 2-class scenarios were tested to find the most discriminable movement type. The linear discriminant analysis achieved on average, higher classification accuracies for both movement detection and classification. The right- and down tongue movements provided the highest and lowest detection accuracy (95.3±4.3% and 91.7±4.8%), respectively. The 4-class classification achieved an accuracy of 62.6±7.2%, while the best 3-class classification (using left, right, and up movements) and 2-class classification (using left and right movements) achieved an accuracy of 75.6±8.4% and 87.7±8.0%, respectively. Using only a combination of the temporal and template feature groups provided further classification accuracy improvements. Presumably, this is because these feature groups utilize the movement-related cortical potentials, which are noticeably different on the left- versus right brain hemisphere for the different movements. This study shows that the cortical representation of the tongue is useful for extracting control signals for multi-class movement detection BCIs.

Synthetic Spatial Foraging With Active Inference in a Geocaching Task

Humans are highly proficient in learning about the environments in which they operate. They form flexible spatial representations of their surroundings that can be leveraged with ease during spatial foraging and navigation. To capture these abilities, we present a deep Active Inference model of goal-directed behavior, and the accompanying belief updating. Active Inference rests upon optimizing Bayesian beliefs to maximize model evidence or marginal likelihood. Bayesian beliefs are probability distributions over the causes of observable outcomes. These causes include an agent's actions, which enables one to treat planning as inference. We use simulations of a geocaching task to elucidate the belief updating-that underwrites spatial foraging-and the associated behavioral and neurophysiological responses. In a geocaching task, the aim is to find hidden objects in the environment using spatial coordinates. Here, synthetic agents learn about the environment via inference and learning (e.g., learning about the likelihoods of outcomes given latent states) to reach a target location, and then forage locally to discover the hidden object that offers clues for the next location.

Response of the Cerebral Cortex to Resistance and Non-resistance Exercise Under Different Trajectories: A Functional Near-Infrared Spectroscopy Study

Background: At present, the effects of upper limb movement are generally evaluated from the level of motor performance. The purpose of this study is to evaluate the response of the cerebral cortex to different upper limb movement patterns from the perspective of neurophysiology.

Method: Thirty healthy adults (12 females, 18 males, mean age 23.9 ± 0.9 years) took resistance and non-resistance exercises under four trajectories (T1: left and right straight-line movement; T2: front and back straight-line movement; T3: clockwise and anticlockwise drawing circle movement; and T4: clockwise and anticlockwise character ⁕ movement). Each movement included a set of periodic motions composed of a 30-s task and a 30-s rest. Functional near-infrared spectroscopy (fNIRS) was used to measure cerebral blood flow dynamics. Primary somatosensory cortex (S1), supplementary motor area (SMA), pre-motor area (PMA), primary motor cortex (M1), and dorsolateral prefrontal cortex (DLPFC) were chosen as regions of interests (ROIs). Activation maps and symmetric heat maps were applied to assess the response of the cerebral cortex to different motion patterns.

Result: The activation of the brain cortex was significantly increased during resistance movement for each participant. Specifically, S1, SMA, PMA, and M1 had higher participation during both non-resistance movement and resistance movement. Compared to non-resistance movement, the resistance movement caused an obvious response in the cerebral cortex. The task state and the resting state were distinguished more obviously in the resistance movement. Four trajectories can be distinguished under non-resistance movement.

Conclusion: This study confirmed that the response of the cerebral motor cortex to different motion patterns was different from that of the neurophysiological level. It may provide a reference for the evaluation of resistance training effects in the future.

KITE-BCI: A brain-computer interface system for functional electrical stimulation therapy

CONTEXT/OBJECTIVE

Integrating brain-computer interface (BCI) technology with functional electrical stimulation therapy (FEST) is an emerging strategy for upper limb motor rehabilitation after spinal cord injury (SCI). Despite promising results, the combined use of these technologies (BCI-FEST) in clinical practice is minimal. To address this issue, we developed KITE-BCI, a BCI system specifically designed for clinical application and integration with dynamic FEST. In this paper, we report its technical features and performance. In addition, we discuss the differences in distributions of the BCI- and therapist-triggered stimulation latencies.

DESIGN

Two single-arm 40-session interventional studies to test the feasibility of BCI-controlled FEST for upper limb motor rehabilitation in individuals with cervical SCI.

SETTING

Rehabilitation programs within the University and Lyndhurst Centres of the Toronto Rehabilitation Institute - University Health Network, Toronto, Canada.

PARTICIPANTS

Five individuals with sub-acute (< 6 months post-injury) SCI at the C4-C5 level, AIS B-D, and three individuals with chronic (> 24 months post-injury) SCI at C4 level, AIS B-C.

OUTCOME MEASURES

We measured BCI setup duration, and to characterize the performance of KITE-BCI, we recorded BCI sensitivity, defined as the percentage of successful BCI activations out of the total number of cued movements.

RESULTS

The overall BCI sensitivities were 74.46% and 79.08% for the sub-acute and chronic groups, respectively. The average KITE-BCI setup duration across the two studies was 11 min and 13 s.

CONCLUSION

KITE-BCI demonstrates a clinically viable single-channel BCI system for integration with FEST resulting in a versatile technology-enhanced upper limb motor rehabilitation strategy after SCI.

Shoulder Flexion Pre-Movement Recognition Through Subject-Specific Brain Regions to Command an Upper Limb Exoskeleton

This work presents two brain-computer interfaces (BCIs) for shoulder pre-movement recognition using: 1) manual strategy for Electroencephalography (EEG) channels selection, and 2) subject-specific channels selection by applying non-negative factorization matrix (NMF). Besides, the proposed BCIs compute spatial features extracted from filtered EEG signals through Riemannian covariance matrices and a linear discriminant analysis (LDA) to discriminate both shoulder pre-movement and rest states. We studied on twenty-one healthy subjects different frequency ranges looking the best frequency band for shoulder pre-movement recognition. As a result, our BCI located automatically EEG channels on the contralateral moved limb, and enhancing the pre-movement recognition (ACC = 71.39 ± 12.68%, κ = 0.43 ± 0.25%). The ability of the proposed BCIs to select specific EEG locations more cortically related to the moved limb could benefit the neuro-rehabilitation process.

A Simplified CNN Classification Method for MI-EEG via the Electrode Pairs Signals

A brain-computer interface (BCI) based on electroencephalography (EEG) can provide independent information exchange and control channels for the brain and the outside world. However, EEG signals come from multiple electrodes, the data of which can generate multiple features. How to select electrodes and features to improve classification performance has become an urgent problem to be solved. This paper proposes a deep convolutional neural network (CNN) structure with separated temporal and spatial filters, which selects the raw EEG signals of the electrode pairs over the motor cortex region as hybrid samples without any preprocessing or artificial feature extraction operations. In the proposed structure, a 5-layer CNN has been applied to learn EEG features, a 4-layer max pooling has been used to reduce dimensionality, and a fully-connected (FC) layer has been utilized for classification. Dropout and batch normalization are also employed to reduce the risk of overfitting. In the experiment, the 4 s EEG data of 10, 20, 60, and 100 subjects from the Physionet database are used as the data source, and the motor imaginations (MI) tasks are divided into four types: left fist, right fist, both fists, and both feet. The results indicate that the global averaged accuracy on group-level classification can reach 97.28%, the area under the receiver operating characteristic (ROC) curve stands out at 0.997, and the electrode pair with the highest accuracy on 10 subjects dataset is FC3-FC4, with 98.61%. The research results also show that this CNN classification method with minimal (2) electrode can obtain high accuracy, which is an advantage over other methods on the same database. This proposed approach provides a new idea for simplifying the design of BCI systems, and accelerates the process of clinical application.

A functional source separation algorithm to enhance error-related potentials monitoring in noninvasive brain-computer interface

BACKGROUND AND OBJECTIVES

An Error related Potential (ErrP) can be noninvasively and directly measured from the scalp through electroencephalography (EEG), as response, when a person realizes they are making an error during a task (as a consequence of a cognitive error performed from the user). It has been shown that ErrPs can be automatically detected with time-discrete feedback tasks, which are widely applied in the Brain-Computer Interface (BCI) field for error correction or adaptation. In this work, a semi-supervised algorithm, namely the Functional Source Separation (FSS), is proposed to estimate a spatial filter for learning the ErrPs and to enhance the evoked potentials.

METHODS

EEG data recorded on six subjects were used to evaluate the proposed method based on FFS algorithm in comparison with the xDAWN algorithm. FSS- and xDAWN-based methods were compared also to the Cz and FCz single channel. Single-trial classification was considered to evaluate the performances of the approaches. (Both the approaches were evaluated on single-trial classification of EEGs.) RESULTS: The results presented using the Bayesian Linear Discriminant Analysis (BLDA) classifier, show that FSS (accuracy 0.92, sensitivity 0.95, specificity 0.81, F1-score 0.95) overcomes the other methods (Cz - accuracy 0.72, sensitivity 0.74, specificity 0.63, F1-score 0.74; FCz - accuracy 0.72, sensitivity 0.75, specificity 0.61, F1-score 0.75; xDAWN - accuracy 0.75, sensitivity 0.79, specificity 0.61, F1-score 0.79) in terms of single-trial classification.

CONCLUSIONS

The proposed FSS-based method increases the single-trial detection accuracy of ErrPs with respect to both single channel (Cz, FCz) and xDAWN spatial filter.

Recognition of EEG Signal Motor Imagery Intention Based on Deep Multi-View Feature Learning

Recognition of motor imagery intention is one of the hot current research focuses of brain-computer interface (BCI) studies. It can help patients with physical dyskinesia to convey their movement intentions. In recent years, breakthroughs have been made in the research on recognition of motor imagery task using deep learning, but if the important features related to motor imagery are ignored, it may lead to a decline in the recognition performance of the algorithm. This paper proposes a new deep multi-view feature learning method for the classification task of motor imagery electroencephalogram (EEG) signals. In order to obtain more representative motor imagery features in EEG signals, we introduced a multi-view feature representation based on the characteristics of EEG signals and the differences between different features. Different feature extraction methods were used to respectively extract the time domain, frequency domain, time-frequency domain and spatial features of EEG signals, so as to made them cooperate and complement. Then, the deep restricted Boltzmann machine (RBM) network improved by t-distributed stochastic neighbor embedding(t-SNE) was adopted to learn the multi-view features of EEG signals, so that the algorithm removed the feature redundancy while took into account the global characteristics in the multi-view feature sequence, reduced the dimension of the multi-visual features and enhanced the recognizability of the features. Finally, support vector machine (SVM) was chosen to classify deep multi-view features. Applying our proposed method to the BCI competition IV 2a dataset we obtained excellent classification results. The results show that the deep multi-view feature learning method further improved the classification accuracy of motor imagery tasks.