Motor Imagery Classification Accuracy Research Articles

Online BCI systems: cross-subject motor imagery classification based on weighted time-domain feature extraction methods

Motor imagery electroencephalogram (MI-EEG) is becoming increasingly important. This paper solves the problem of online signal recognition for motor imagery across subjects by finding common features across multiple subjects to improve the generality of the classification model. We analysed the EEG data from left/right-hand motor imagery of eight subjects and proposed a weighted time-domain (WTD) feature extraction method based on a weighted channel screening method. The classification model constructed by combining this feature extraction method with the support vector machine (SVM) classification method was faster in classification and achieved good cross-subject classification accuracy (The average offline classification accuracy was 91.39%). In this paper, an online control system for asynchronous brain-controlled wheelchairs was built with good performance. The online average motor imagery classification accuracy was 81.67%, and the average response time was 1.36s. This method contributes to bringing the online Brain-computer interface (BCI) system out of the laboratory and into wider application.

Short-Interval Priming Effects: An EEG Study of Action Observation on Motor Imagery

Recent studies revealed that setting motor priming prior to main intervention can improve rehabilitation outcomes. However, existing priming strategies generally take dozens of minutes, leading to reduced effective treatment time and rehabilitation prescription flexibility. Priming individual training trial with a short-term (a few seconds) task may be a promising alternative. The current study focused on investigating short-term priming effects via an electroencephalogram (EEG) study of action observation (AO) on motor imagery (MI). Six healthy adults were recruited to perform MI of left or right elbow flexion-extension primed by AO of the same action. The control condition replaced AO priming with directional arrows indicating different elbows. We employed the event-related desynchronization (ERD) of µ rhythm (8-13 Hz) and the classification accuracy of MI to evaluate short-term priming effects. The obtained results revealed that both ERD intensity and classification accuracy of AO-primed MI were greater than those of arrow cue-guided MI, supporting short-term priming as a feasible adjuvant technique to boost the effect of main intervention.

Optimal Deep Learning-Based Recognition Model for EEG Enabled Brain-Computer Interfaces Using Motor-Imagery

Abstract Brain-Computer Interfaces (BCIs) facilitate the translation of brain activity into actionable commands and act as a crucial link between the human brain and the external environment. Electroencephalography (EEG)-based BCIs, which focus on motor imagery, have emerged as an important area of study in this domain. They are used in neurorehabilitation, neuroprosthetics, and gaming, among other applications. Optimal Deep Learning-Based Recognition for EEG Signal Motor Imagery (ODLR-EEGSM) is a novel approach presented in this article that aims to improve the recognition of motor imagery from EEG signals. The proposed method includes several crucial stages to improve the precision and effectiveness of EEG-based motor imagery recognition. The pre-processing phase starts with the Variation Mode Decomposition (VMD) technique, which is used to improve EEG signals. The EEG signals are decomposed into different oscillatory modes by VMD, laying the groundwork for subsequent feature extraction. Feature extraction is a crucial component of the ODLR-EEGSM method. In this study, we use Stacked Sparse Auto Encoder (SSAE) models to identify significant patterns in the pre-processed EEG data. Our approach is based on the classification model using Deep Wavelet Neural Network (DWNN) optimized with Chaotic Dragonfly Algorithm (CDFA). CDFA optimizes the weight and bias values of the DWNN, significantly improving the classification accuracy of motor imagery. To evaluate the efficacy of the ODLR-EEGSM method, we use benchmark datasets to perform rigorous performance validation. The results show that our approach outperforms current methods in the classification of EEG motor imagery, confirming its promising performance. This study has the potential to make brain-computer interface applications in various fields more accurate and efficient, and pave the way for brain-controlled interactions with external systems and devices.

A cross-session motor imagery classification method based on Riemannian geometry and deep domain adaptation

Recently, more and more studies have begun to use deep learning to decode and classify EEG signals. The use of deep learning has led to an increase in the classification accuracy of motor imagery (MI), but the problem of taking a long time to calibrate in brain–computer interface (BCI) applications has not been solved. To address this problem, we propose a novel Riemannian geometry and deep domain adaptation network (RGDDANet) for MI classification. Specifically, two one-dimensional convolutions are designed to extract temporal and spatial features from the EEG signals, and then the spatial covariance matrices are utilized to map the extracted features to Riemannian manifolds for processing. In order to align the source and target features’ distributions on the Riemannian manifold, we propose a Symmetric Positive Definite (SPD) matrix mean discrepancy loss (SMMDL) to minimize the distance between two domains. To analyze the feasibility of the method, we conducted extensive experiments on BCIC IV 2a and BCIC IV 2b datasets, respectively, and the results showed that the proposed method achieved better performance than some state-of-the-art methods.

Classification of Motor Imagery Using Trial Extension in Spatial Domain with Rhythmic Components of EEG

Electrical activities of the human brain can be recorded with electroencephalography (EEG). To characterize motor imagery (MI) tasks for brain–computer interface (BCI) implementation is an easy and cost-effective tool. The MI task is represented by a short-time trial of multichannel EEG. In this paper, the signal of each channel of raw EEG is decomposed into a finite set of narrowband signals using a Fourier-transformation-based bandpass filter. Rhythmic components of EEG are represented by each of the narrowband signals that characterize the brain activities related to MI tasks. The subband signals are arranged to extend the dimension of the EEG trial in the spatial domain. The spatial features are extracted from the set of extended trials using a common spatial pattern (CSP). An optimum number of features are employed to classify the motor imagery tasks using an artificial neural network. An integrated approach with full-band and narrowband signals is implemented to derive discriminative features for MI classification. In addition, the subject-dependent parameter optimization scheme enhances the performance of the proposed method. The performance evaluation of the proposed method is obtained using two publicly available benchmark datasets (Dataset I and Dataset II). The experimental results in terms of classification accuracy (93.88% with Dataset I and 91.55% with Dataset II) show that it performs better than the recently developed algorithms. The enhanced MI classification accuracy is very much applicable in BCI implementation.

Priming cross-session motor imagery classification with a universal deep domain adaptation framework

Electroencephalogram (EEG) based motor imagery (MI) brain–computer interfaces (BCI) are widely used in applications related to rehabilitation and external device control. However, due to the non-stationary and low signal-to-noise ratio characteristics of EEG, classifying motor imagery tasks of the same participant from different recording sessions is generally challenging. Whether the classification accuracy of cross-session MI can be improved from the perspective of domain adaptation is a question worth verifying. In this paper, we propose a Siamese deep domain adaptation (SDDA) framework for cross-session MI classification based on mathematical models in domain adaptation theory. The SDDA framework primarily consists of three components: a novel preprocessing method based on domain-invariant features, a maximum mean discrepancy (MMD) loss for aligning source and target domain embedding features, and an improved cosine-based center loss designed to suppress the influence of noise and outliers on the neural network. The SDDA framework has been validated with two classic and popular convolutional neural networks (EEGNet and ConvNet) from BCI research field in two MI EEG public datasets (BCI Competition IV IIA, IIB). Compared with the vanilla EEGNet and ConvNet, the SDDA framework improves the MI classification accuracy by 10.49%, 7.60% respectively in IIA dataset, and 4.59%, 3.35% in IIB dataset. The SDDA not only significantly improves the classification performance of the vanilla networks but also surpasses state-of-the-art transfer learning methods, making it a superior and user-friendly approach for MI classification.

EEG-CDILNet: a lightweight and accurate CNN network using circular dilated convolution for motor imagery classification.

Objective.The combination of the motor imagery (MI) electroencephalography (EEG) signals and deep learning-based methods is an effective way to improve MI classification accuracy. However, deep learning-based methods often need too many trainable parameters. As a result, the trade-off between the network decoding performance and computational cost has always been an important challenge in the MI classification research.Approach.In the present study, we proposed a new end-to-end convolutional neural network (CNN) model called the EEG-circular dilated convolution (CDIL) network, which takes into account both the lightweight model and the classification accuracy. Specifically, the depth-separable convolution was used to reduce the number of network parameters and extract the temporal and spatial features from the EEG signals. CDIL was used to extract the time-varying deep features that were generated in the previous stage. Finally, we combined the features extracted from the two stages and used the global average pooling to further reduce the number of parameters, in order to achieve an accurate MI classification. The performance of the proposed model was verified using three publicly available datasets.Main results.The proposed model achieved an average classification accuracy of 79.63% and 94.53% for the BCIIV2a and HGD four-classification task, respectively, and 87.82% for the BCIIV2b two-classification task. In particular, by comparing the number of parameters, computation and classification accuracy with other lightweight models, it was confirmed that the proposed model achieved a better balance between the decoding performance and computational cost. Furthermore, the structural feasibility of the proposed model was confirmed by ablation experiments and feature visualization.Significance.The results indicated that the proposed CNN model presented high classification accuracy with less computing resources, and can be applied in the MI classification research.

Improving single-hand open/close motor imagery classification by error-related potentials correction

ObjectiveThe ability of a brain-computer interface (BCI) to classify brain activity in electroencephalograms (EEG) during motor imagery (MI) tasks is an important performance indicator. Because the cortical regions that drive the single-handed open and closed tasks overlap, it is difficult to classify the EEG signals during executing both tasks. ApproachThe addition of special EEG features can improve the accuracy of classifying single-hand open and closed tasks. In this work, we designed a hybrid BCI paradigm based on error-related potentials (ErrP) and motor imagery (MI) and proposed a strategy to correct the classification results of MI by using ErrP information. The ErrP and MI features of EEG data from 11 subjects were superimposed. Main resultsThe corrected strategy improved the classification accuracy of single-hand open/close MI tasks from 52.3% to 73.7%, an increase of approximately 21%. SignificanceOur hybrid BCI paradigm improves the classification accuracy of single-hand MI by adding ErrP information, which provides a new approach for improving the classification performance of BCI.

Motor intent recognition of multi-feature fusion EEG signals by UMAP algorithm.

The key to the analysis of electroencephalogram (EEG) signals lies in the extraction of effective features from the raw EEG signals, which can then be utilized to augment the classification accuracy of motor imagery (MI) applications in brain-computer interface (BCI). It can be argued that the utilization of features from multiple domains can be a more effective approach to feature extraction for MI pattern classification, as it can provide a more comprehensive set of information that the traditional single feature extraction method may not be able to capture. In this paper, a multi-feature fusion algorithm based on uniform manifold approximate and projection (UMAP) is proposed for motor imagery EEG signals. The brain functional network and common spatial pattern (CSP) are initially extracted as features. Subsequently, UMAP is utilized to fuse the extracted multi-domain features to generate low-dimensional features with improved discriminative capability. Finally, the k-nearest neighbor (KNN) classifier is applied in a lower dimensional space. The proposed method is evaluated using left-right hand EEG signals, and achieved the average accuracy of over 92%. The results indicate that, compared with single-domain-based feature extraction methods, multi-feature fusion EEG signal classification based on the UMAP algorithm yields superior classification and visualization performance. Feature extraction and fusion based on UMAP algorithm of left-right hand motor imagery.

A deep domain adaptation framework with correlation alignment for EEG-based motor imagery classification

It is impractical to collect sufficient and well-labeled EEG data in Brain-computer interface because of the time-consuming data acquisition and costly annotation. Conventional classification methods reusing EEG data from different subjects and time periods (across domains) significantly decrease the classification accuracy of motor imagery. In this paper, we propose a deep domain adaptation framework with correlation alignment (DDAF-CORAL) to solve the problem of distribution divergence for motor imagery classification across domains. Specifically, a two-stage framework is adopted to extract deep features for raw EEG data. The distribution divergence caused by subjected-related and time-related variations is further minimized by aligning the covariance of the source and target EEG feature distributions. Finally, the classification loss and adaptation loss are optimized simultaneously to achieve sufficient discriminative classification performance and low feature distribution divergence. Extensive experiments on three EEG datasets demonstrate that our proposed method can effectively reduce the distribution divergence between the source and target EEG data. The results show that our proposed method delivers outperformance (an average classification accuracy of 92.9% for within-session, an average kappa value of 0.761 for cross-session, and an average classification accuracy of 83.3% for cross-subject) in two-class classification tasks compared to other state-of-the-art methods.

A signal prediction-based method for motor imagery EEG classification

Motor imagery (MI)-based brain-computer interface (BCI) usually uses EEG signals recorded from numerous scalp electrodes for accurate MI classification, which requires long preparation time and high cost, hindering its practical application. A novel method based on signal prediction was proposed in this study to achieve high MI classification accuracy by using EEG recorded from only a small number of electrodes. The signal prediction model was built by using regression technique to estimate the full-channel EEG signals (e.g., recorded from 19, 22, and 29 electrodes) from the few-channel EEG signals on the central region, and then the predicted full-channel EEG signals were used for MI feature extraction and classification. The proposed methods based on multiple linear regression, ridge regression and extreme learning machine regression were compared with the traditional methods that directly used recorded few-channel and full-channel EEG signals, respectively, on three public EEG datasets. The proposed method significantly outperformed the traditional method that directly used few-channel EEG, and was comparable to or even superior to the traditional method that directly used full-channel EEG. The proposed method can accurately predict full-channel EEG and achieve high classification accuracy with only a small number of electrodes with fixed locations, which has great potential in building MI-based BCIs that meet the requirements of practical applications.

A study of action difference on motor imagery based on delayed matching posture task

Objective. Motor imagery (MI)-based brain-computer interfaces (BCIs) provide an additional control pathway for people by decoding the intention of action imagination. The way people imagine greatly affects MI-BCI performance. Action itself is one of the factors that influence the way people imagine. Whether the different actions cause a difference in the MI performance is unknown. What is more important is how to manifest this action difference in the process of imagery, which has the potential to guide people to use their individualized actions to imagine more effectively. Approach. To explore action differences, this study proposes a novel paradigm named as action observation based delayed matching posture task. Ten subjects are required to observe, memorize, match, and imagine three types of actions (cutting, grasping and writing) given by visual images or videos, to accomplish the phases of encoding, retrieval and reinforcement of MI. Event-related potential (ERP), MI features, and classification accuracy of the left or the right hand are used to evaluate the effect of the action difference on the MI difference. Main results. Action differences cause different feature distributions, resulting in that the accuracy with high event-related (de)synchronization (ERD/ERS) is 27.75% higher than the ones with low ERD/ERS (p < 0.05), which indicates that the action difference has impact on the MI difference and the BCI performance. In addition, significant differences in the ERP amplitudes exists among the three actions: the amplitude of P300-N200 potential reaches 9.28 μV of grasping, 5.64 μV and 5.25 μV higher than the cutting and the writing, respectively (p < 0.05). Significance. The ERP amplitudes derived from the supplementary motor area shows positive correlation to the MI classification accuracy, implying that the ERP might be an index of the MI performance when the people is faced with action selection. This study demonstrates that the MI difference is related to the action difference, and can be manifested by the ERP, which is important for improving MI training by selecting suitable action; the relationship between the ERP and the MI provides a novel index to find the suitable action to set up an individualized BCI and improve the performance further.

Enhancement of motor imagery training efficiency by an online adaptive training paradigm integrated with error related potential

Objective. Motor imagery (MI) is a process of autonomously modulating the motor area to rehearse action mentally without actual execution. Based on the neuroplasticity of the cerebral cortex, MI can promote the functional rehabilitation of the injured cerebral cortex motor area. However, it usually takes several days to a few months to train individuals to acquire the necessary MI ability to control rehabilitation equipment in current studies, which greatly limits the clinical application of rehabilitation training systems based on the MI brain–computer interface (BCI). Approach. A novel MI training paradigm combined with the error related potential (ErrP) is proposed, and online adaptive training of the MI classifier was performed using ErrP. ErrP is used to correct the output of the MI classification to obtain a higher accuracy of kinesthetic feedback based on the imagination intention of subjects while generating simulated labels for MI online adaptive training. In this way, we improved the MI training efficiency. Thirteen subjects were randomly divided into an experimental group using the proposed paradigm and a control group using the traditional MI training paradigm to participate in six MI training experiments. Main results. The proposed paradigm enabled the experimental group to obtain a higher event-related desynchronization modulation level in the contralateral brain region compared with the control group and 69.76% online classification accuracy of MI after three MI training experiments. The online classification accuracy reached 72.76% and the whole system recognized the MI intention of the subjects with an online accuracy of 82.61% after six experiments. Significance. Compared with the conventional unimodal MI training strategy, the proposed approach enables subjects to use the MI-BCI based system directly and achieve a better performance after only three training experiments with training left and right hands simultaneously. This greatly improves the usability of the MI-BCI-based rehabilitation system and makes it more convenient for clinical use.

A novel residual shrinkage block-based convolutional neural network for improving the recognition of motor imagery EEG signals

PurposeRecently, the convolutional neural network (ConvNet) has a wide application in the classification of motor imagery EEG signals. However, the low signal-to-noise electroencephalogram (EEG) signals are collected under the interference of noises. However, the conventional ConvNet model cannot directly solve this problem. This study aims to discuss the aforementioned issues.Design/methodology/approachTo solve this problem, this paper adopted a novel residual shrinkage block (RSB) to construct the ConvNet model (RSBConvNet). During the feature extraction from EEG signals, the proposed RSBConvNet prevented the noise component in EEG signals, and improved the classification accuracy of motor imagery. In the construction of RSBConvNet, the author applied the soft thresholding strategy to prevent the non-related motor imagery features in EEG signals. The soft thresholding was inserted into the residual block (RB), and the suitable threshold for the current EEG signals distribution can be learned by minimizing the loss function. Therefore, during the feature extraction of motor imagery, the proposed RSBConvNet de-noised the EEG signals and improved the discriminative of classification features.FindingsComparative experiments and ablation studies were done on two public benchmark datasets. Compared with conventional ConvNet models, the proposed RSBConvNet model has obvious improvements in motor imagery classification accuracy and Kappa coefficient. Ablation studies have also shown the de-noised abilities of the RSBConvNet model. Moreover, different parameters and computational methods of the RSBConvNet model have been tested on the classification of motor imagery.Originality/valueBased on the experimental results, the RSBConvNet constructed in this paper has an excellent recognition accuracy of MI-BCI, which can be used for further applications for the online MI-BCI.

Learning Optimal Time-Frequency-Spatial Features by the CiSSA-CSP Method for Motor Imagery EEG Classification.

The common spatial pattern (CSP) is a popular method in feature extraction for motor imagery (MI) electroencephalogram (EEG) classification in brain-computer interface (BCI) systems. However, combining temporal and spectral information in the CSP-based spatial features is still a challenging issue, which greatly affects the performance of MI-based BCI systems. Here, we propose a novel circulant singular spectrum analysis embedded CSP (CiSSA-CSP) method for learning the optimal time-frequency-spatial features to improve the MI classification accuracy. Specifically, raw EEG data are first segmented into multiple time segments and spectrum-specific sub-bands are further derived by CiSSA from each time segment in a set of non-overlapping filter bands. CSP features extracted from all time-frequency segments contain more sufficient time-frequency-spatial information. An experimental study was implemented on the publicly available EEG dataset (BCI Competition III dataset IVa) and a self-collected experimental EEG dataset to validate the effectiveness of the CiSSA-CSP method. Experimental results demonstrate that discriminative and robust features are extracted effectively. Compared with several state-of-the-art methods, the proposed method exhibited optimal accuracies of 96.6% and 95.2% on the public and experimental datasets, respectively, which confirms that it is a promising method for improving the performance of MI-based BCIs.

EEG channel selection based on sequential backward floating search for motor imagery classification.

Brain-computer interfaces (BCIs) based on motor imagery (MI) utilizing multi-channel electroencephalogram (EEG) data are commonly used to improve motor function of people with motor disabilities. EEG channel selection can enhance MI classification accuracy by selecting informative channels, accordingly reducing redundant information. The sequential backward floating search (SBFS) approach has been considered as one of the best feature selection methods. In this paper, SBFS is first implemented to select the optimal EEG channels in MI-BCI. Further, to reduce the time complexity of SBFS, the modified SBFS is proposed and applied to left and right hand MI tasks. In the modified SBFS, based on the map of EEG channels at the scalp, the symmetrical channels are selected as channel pairs and acceleration is thus realized by removing or adding multiple channels in each iteration. Extensive experiments were conducted on four public BCI datasets. Experimental results show that the SBFS achieves significantly higher classification accuracy (p < 0.001) than using all channels and conventional MI channels (i.e., C3, C4, and Cz). Moreover, the proposed method outperforms the state-of-the-art selection methods.

Wearable Supernumerary Robotic Limb System Using a Hybrid Control Approach Based on Motor Imagery and Object Detection

Motor disorder of upper limbs has seriously affected the daily life of the patients with hemiplegia after stroke. We developed a wearable supernumerary robotic limb (SRL) system using a hybrid control approach based on motor imagery (MI) and object detection for upper-limb motion assistance. SRL system included an SRL hardware subsystem and a hybrid control software subsystem. The system obtained the patient's motion intention through MI electroencephalogram (EEG) recognition method based on graph convolutional network (GCN) and gated recurrent unit network (GRU) to control the left and right movements of SRL, and the object detection technology was used together for a quick grasp of target objects to compensate for the disadvantages when using MI EEG alone like fewer control instructions and lower control efficiency. Offline training experiment was designed to obtain subjects' MI recognition models and evaluate the feasibility of the MI EEG recognition method; online control experiment was designed to verify the effectiveness of our wearable SRL system. The results showed that the proposed MI EEG recognition method (GCN+GRU) could effectively improve the MI classification accuracy (90.04% ± 2.36 %) compared with traditional methods; all subjects were able to complete the target object grasping tasks within 23 seconds by controlling the SRL, and the highest average grasping success rate achieved 90.67% in bag grasping task. The SRL system can effectively assist people with upper-limb motor disorder to perform upper-limb tasks in daily life by natural human-robot interaction, and improve their ability of self-help and enhance their confidence of life.

Enhancement of capability for motor imagery using vestibular imbalance stimulation during brain computer interface

Objective. Motor imagery (MI), based on the theory of mirror neurons and neuroplasticity, can promote motor cortical activation in neurorehabilitation. The strategy of MI based on brain-computer interface (BCI) has been used in rehabilitation training and daily assistance for patients with hemiplegia in recent years. However, it is difficult to maintain the consistency and timeliness of receiving external stimulation to neural activation in most subjects owing to the high variability of electroencephalogram (EEG) representation across trials/subjects. Moreover, in practical application, MI-BCI cannot highly activate the motor cortex and provide stable interaction owing to the weakness of the EEG feature and lack of an effective mode of activation. Approach. In this study, a novel hybrid BCI paradigm based on MI and vestibular stimulation motor imagery (VSMI) was proposed to enhance the capability of feature response for MI. Twelve subjects participated in a group of controlled experiments containing VSMI and MI. Three indicators, namely, activation degree, timeliness, and classification accuracy, were adopted to evaluate the performance of the task. Main results. Vestibular stimulation could significantly strengthen the suppression of α and β bands of contralateral brain regions during MI, that is, enhance the activation degree of the motor cortex (p< 0.01). Compared with MI, the timeliness of EEG feature-response achieved obvious improvements in VSMI experiments. Moreover, the averaged classification accuracy of VSMI and MI was 80.56% and 69.38%, respectively. Significance. The experimental results indicate that specific vestibular activity contributes to the oscillations of the motor cortex and has a positive effect on spontaneous imagery, which provides a novel MI paradigm and enables the preliminary exploration of sensorimotor integration of MI.

EEG-based motor imagery classification using convolutional neural networks with local reparameterization trick

ObjectivesDeep learning (DL) method has emerged as a powerful tool in studying the behavior of Electroencephalogram (EEG)-based motor imagery (MI). Although prospective studies have demonstrated promising performance, most of these studies have been affected by the lack of research between groups and individual subjects, and the accuracy of MI classification still has room for improvement. Due to the inter-individual variability in the EEG classification, enhancing the adaptability and robustness between different individuals is especially critical. Methods & experimentsWe developed a novel DL model based on the EEG signals to improve MI classification performance by introducing the local reparameterization trick into convolutional neural networks (LRT-CNN). 109 subjects from PhysioNet Dataset were used to test the proposed model. Firstly, a global classifier was evaluated by four groups. Secondly, individual variability was examined by testing individual subjects. ResultsThe classification accuracy of global classifier in 20 subjects, 50 subjects, 80 subjects, and 109 subjects are 93.86%, 98.94%, 93.04%, and 92.41%, respectively. The maximum classification accuracy of one individual subject is 99.79%, which is better than the state-of-the-art method and proves the proposed method can handle the challenge of individual variability. ConclusionWe conclude that introducing the local reparameterization trick into convolutional neural networks can significantly improve the accuracy of the MI tasks based on the EEG signals without any complicated and tedious feature engineering works. Besides, encouraging results were obtained both between groups (multiple subjects) and on a single subject. ImplicationThe experimental results add to the rapidly expanding field of brain science and contribute to our understanding of applying the DL method to address EEG-based classification problems (not limited to MI classification issues).

A Novel Two-Stage Refine Filtering Method for EEG-Based Motor Imagery Classification.

Cerebral stroke is a common disease across the world, and it is a promising method to recognize the intention of stroke patients with the help of brain–computer interface (BCI). In the field of motor imagery (MI) classification, appropriate filtering is vital for feature extracting of electroencephalogram (EEG) signals and consequently influences the accuracy of MI classification. In this case, a novel two-stage refine filtering method was proposed, inspired by Gradient-weighted Class Activation Mapping (Grad-CAM), which uses the gradients of any target concept flowing into the final convolutional layer to highlight the important part of training data for predicting the concept. In the first stage, MI classification was carried out and then the frequency band to be filtered was calculated according to the Grad-CAM of the MI classification results. In the second stage, EEG was filtered and classified for a higher classification accuracy. To evaluate the filtering effect, this method was applied to the multi-branch neural network proposed in our previous work. Experiment results revealed that the proposed method reached state-of-the-art classification kappa value levels and acquired at least 3% higher kappa values than other methods This study also proposed some promising application scenarios with this filtering method.