Dataset 2a Research Articles

Domain Generalization through Latent Distribution Exploration for Motor Imagery EEG Classification

Electroencephalography (EEG)-based Motor Imagery (MI) brain-computer interface (BCI) systems play essential roles in motor function rehabilitation for patients with post-stroke. Existing neural networks for decoding MI EEG face challenges due to nonstationary characteristics and subject-specific variations of EEG data. To address these challenges and improve generalization performance, this study proposes a domain generalization (DG) model that eliminates the need for user-specific calibration in real-life applications. Specifically, the proposed model comprises two branches: the first branch applies several independent decision-making networks to decode and classify subjects’ motor intentions, while the second branch adaptively assigns weights to classification results and fuses them into a comprehensive decision. Both branches utilize EEGNet and ShallowConvNet to extract time-frequency-spatial features. By implementing multiple classification networks, the model can learn a broad range of data distributions from source subjects, which contributes to improved generalization performance on target subjects. The proposed EEG-DG framework was evaluated on BCI Competition IV Dataset 2a, 2b and PhysioNet. Results show that the proposed framework significantly enhances the classification performance of MI EEG, outperforming several state-of-the-art models on all three datasets, underlining its superior efficacy in real-world scenarios and exceptional generalization performance. The source code can be accessed at https://github.com/DrugLover/Multibranch-DG-EEG.

Compact convolutional transformer for subject-independent motor imagery EEG-based BCIs

Motor imagery electroencephalography (EEG) analysis is crucial for the development of effective brain-computer interfaces (BCIs), yet it presents considerable challenges due to the complexity of the data and inter-subject variability. This paper introduces EEGCCT, an application of compact convolutional transformers designed specifically to improve the analysis of motor imagery tasks in EEG. Unlike traditional approaches, EEGCCT model significantly enhances generalization from limited data, effectively addressing a common limitation in EEG datasets. We validate and test our models using the open-source BCI Competition IV datasets 2a and 2b, employing a Leave-One-Subject-Out (LOSO) strategy to ensure subject-independent performance. Our findings demonstrate that EEGCCT not only outperforms conventional models like EEGNet in standard evaluations but also achieves better performance compared to other advanced models such as Conformer, Hybrid s-CViT, and Hybrid t-CViT, while utilizing fewer parameters and achieving an accuracy of 70.12%. Additionally, the paper presents a comprehensive ablation study that includes targeted data augmentation, hyperparameter optimization, and architectural improvements.

Diffusion models-based motor imagery EEG sample augmentation via mixup strategy

Deep representation learning has been widely explored for decoding motor imagery electroencephalogram (MI-EEG) to build EEG-tailored brain-computer interfaces. Due to the labor-intensive and time-consuming recording to MI-EEG, recently BCIs were suffered from sample scarcity, especially for the limitation of sample diversity. To solve this issue, we proposed a novel sample augmentation method, Diffusion models-based EEG Sample Augmentation method via Mixup strategy (DESAM), to solve sample scarcity and improve sample diversity for both multivariate time-series and time–frequency images forms. To reduce the characteristics conflicts brought by the augmented samples, a mixup strategy with data weighting was proposed for both forms. To validate the efficacy of sample augmentation by DESAM, we conducted comparative experiments on three publicly available MI-EEG datasets, and reported the average decoding accuracies and Kappa values on augmented samples by various deep learning models. For the BCI Competition IV datasets 2a, 2b, and 1, the two forms of DESAM sample augmentations achieved improvements for both accuracies and Kappa values. Ablation studies have demonstrated the necessity and significance of our DESAM method, and it makes a possible manner to solve sample scarcity issue of MI-EEG.

An SCA-based classifier for motor imagery EEG classification

Efficient and accurate multi-class classification of electroencephalogram (EEG) signals poses a significant challenge in the development of motor imagery-based brain–computer interface (MI-BCI). Drawing inspiration from the sine cosine algorithm (SCA), a widely employed swarm intelligence algorithm for optimization problems, we proposed a novel population-based classification algorithm for EEG signals in this article. To fully leverage the characteristics contained in EEG signals, multi-scale sub-signals were constructed in terms of temporal windows and spectral bands simultaneously, and the common spatial pattern (CSP) features were extracted from each sub-signal. Subsequently, we integrated the multi-center optimal vectors mechanism into the classical SCA, resulting in the development of a multi-center SCA (MCSCA) classifier. During the classification stage, the label was assigned to the test trials by evaluating the Euclidean distance between their feature vectors and each optimal vector in MCSCA. Additionally, the weights of feature vectors were exploited to select the sub-signal of specific temporal windows and spectral bands for feature reduction, thereby declining computational effort and eliminating data redundancy. To validate the performance of the MCSCA classifier, we conducted four-class classification experiments using the BCI Competition IV dataset 2a, achieving an average classification accuracy of 71.89%. The experimental results show that the proposed algorithm offers a novel and effective approach for EEG classification.

Enhancing EEG-Based MI-BCIs with Class-Specific and Subject-Specific Features Detected by Neural Manifold Analysis.

This paper presents an innovative approach leveraging Neuronal Manifold Analysis of EEG data to identify specific time intervals for feature extraction, effectively capturing both class-specific and subject-specific characteristics. Different pipelines were constructed and employed to extract distinctive features within these intervals, specifically for motor imagery (MI) tasks. The methodology was validated using the Graz Competition IV datasets 2A (four-class) and 2B (two-class) motor imagery classification, demonstrating an improvement in classification accuracy that surpasses state-of-the-art algorithms designed for MI tasks. A multi-dimensional feature space, constructed using NMA, was built to detect intervals that capture these critical characteristics, which led to significantly enhanced classification accuracy, especially for individuals with initially poor classification performance. These findings highlight the robustness of this method and its potential to improve classification performance in EEG-based MI-BCI systems.

Motor imagery EEG signal classification based on deformable convolution v3 and adaptive spatial attention mechanism

This paper is dedicated to solving the issue of insufficient feature extraction and declining model performance in the classification of motor imagery electroencephalogram (EEG) signals. We propose an enhanced ResNet18 neural network model and introduce two pivotal innovations at the model level. Firstly, we present the deformable convolution v3 (DCNv3), an improved convolutional structure that overcomes the limitations of the original deformable convolution. Through unbiased design with center sampling points, DCNv3 achieves more precise feature localization, enhances the model’s representation capability for complex structures, and reduces the risk of overfitting. Simultaneously, it comprehensively captures features in effective information regions, thereby strengthening the model’s generalization ability. Furthermore, this enhancement design also improves computational efficiency, reducing computational and parameter burdens. Secondly, we introduce a novel Adaptive Spatial Attention Mechanism (ASAM) aimed at precisely capturing feature correlations, thereby avoiding information loss. We conducted extensive comparative experiments on publicly available datasets BCI IV dataset 2a, BCI IV dataset 2b, and BCI III dataset 3a, achieving accuracy rates of 91.32%, 92.3%, and 90.47% respectively. Compared to other methods, our approach not only significantly improves the average classification accuracy but also demonstrates remarkable generalization capabilities. These results validate the effectiveness and stability of our method, thereby better meeting the practical application requirements of brain-computer interface technology.

Multiclass motor imagery classification with Riemannian geometry and temporal-spectral selection.

Motor imagery (MI) based brain-computer interfaces (BCIs) decode the users' intentions from electroencephalography (EEG) to achieve information control and interaction between the brain and external devices. In this paper, firstly, we apply Riemannian geometry to the covariance matrix extracted by spatial filtering to obtain robust and distinct features. Then, a multiscale temporal-spectral segmentation scheme is developed to enrich the feature dimensionality. In order to determine the optimal feature configurations, we utilize a linear learning-based temporal window and spectral band (TWSB) selection method to evaluate the feature contributions, which efficiently reduces the redundant features and improves the decoding efficiency without excessive loss of accuracy. Finally, support vector machines are used to predict the classification labels based on the selected MI features. To evaluate the performance of our model, we test it on the publicly available BCI Competition IV dataset 2a and 2b. The results show that the method has an average accuracy of 79.1% and 83.1%, which outperforms other existing methods. Using TWSB feature selection instead of selecting all features improves the accuracy by up to about 6%. Moreover, the TWSB selection method can effectively reduce the computational burden. We believe that the framework reveals more interpretable feature information of motor imagery EEG signals, provides neural responses discriminative with high accuracy, and facilitates the performance of real-time MI-BCI.

Classification of Motor Imagery EEG signals using high resolution time-frequency representations and convolutional neural network

A Motor Imagery (MI) based Brain Computer Interface (BCI) system aims to provide neuro-rehabilitation for the motor disabled people and patients with brain injuries (e.g., stroke patients) etc. The aim of this work is to classify the left and right hand MI tasks by utilizing the occurrence of event related desynchronization and synchronization (ERD\\ERS) in the Electroencephalogram (EEG) during these tasks. This study proposes to use a set of Complex Morlet Wavelets (CMW) having frequency dependent widths to generate high-resolution time-frequency representations (TFR) of the MI EEG signals present in the channels C3 and C4. A novel method for the selection of the value of number of cycles relative to the center frequency of the CMW is studied here for extracting the MI task features. The generated TFRs are given as input to a Convolutional neural network (CNN) for classifying them into left or right hand MI tasks. The proposed framework attains a classification accuracy of 82.2% on the BCI Competition IV dataset 2a, showing that the TFRs generated in this work give a higher classification accuracy than the baseline methods and other existing algorithms.

Artifact removal and motor imagery classification in EEG using advanced algorithms and modified DNN

This paper presents an advanced approach for EEG artifact removal and motor imagery classification using a combination of Four Class Iterative Filtering and Filter Bank Common Spatial Pattern Algorithm with a Modified Deep Neural Network (DNN) classifier. The research aims to enhance the accuracy and reliability of BCI systems by addressing the challenges posed by EEG artifacts and complex motor imagery tasks.The methodology begins by introducing FCIF, a novel technique for ocular artifact removal, utilizing iterative filtering and filter banks. FCIF's mathematical formulation allows for effective artifact mitigation, thereby improving the quality of EEG data. In tandem, the FC-FBCSP algorithm is introduced, extending the Filter Bank Common Spatial Pattern approach to handle four-class motor imagery classification. The Modified DNN classifier enhances the discriminatory power of the FC-FBCSP features, optimizing the classification process.The paper showcases a comprehensive experimental setup, featuring the utilization of BCI Competition IV Dataset 2a & 2b. Detailed preprocessing steps, including filtering and feature extraction, are presented with mathematical rigor. Results demonstrate the remarkable artifact removal capabilities of FCIF and the classification prowess of FC-FBCSP combined with the Modified DNN classifier. Comparative analysis highlights the superiority of the proposed approach over baseline methods and the method achieves the mean accuracy of 98.575%.

EEG rhythm separation and time-frequency analysis of fast multivariate empirical mode decomposition for motor imagery BCI.

Motor imagery electroencephalogram (EEG) is widely employed in brain-computer interface (BCI) systems. As a time-frequency analysis method for nonlinear and non-stationary signals, multivariate empirical mode decomposition (MEMD) and its noise-assisted version (NA-MEMD) has been widely used in the preprocessing step of BCI systems for separating EEG rhythms corresponding to specific brain activities. However, when applied to multichannel EEG signals, MEMD or NA-MEMD often demonstrate low robustness to noise and high computational complexity. To address these issues, we have explored the advantages of our recently proposed fast multivariate empirical mode decomposition (FMEMD) and its noise-assisted version (NA-FMEMD) for analyzing motor imagery data. We emphasize that FMEMD enables a more accurate estimation of EEG frequency information and exhibits a more noise-robust decomposition performance with improved computational efficiency. Comparative analysis with MEMD on simulation data and real-world EEG validates the above assertions. The joint average frequency measure is employed to automatically select intrinsic mode functions that correspond to specific frequency bands. Thus, FMEMD-based classification architecture is proposed. Using FMEMD as a preprocessing algorithm instead of MEMD can improve the classification accuracy by 2.3% on the BCI Competition IV dataset. On the Physiobank Motor/Mental Imagery dataset and BCI Competition IV Dataset 2a, FMEMD-based architecture also attained a comparable performance to complex algorithms. The results indicate that FMEMD proficiently extracts feature information from small benchmark datasets while mitigating dimensionality constraints resulting from computational complexity. Hence, FMEMD or NA-FMEMD can be a powerful time-frequency preprocessing method for BCI.

Direct lingam and visibility graphs for analyzing brain connectivity in BCI.

The brain-computer interface (BCI) is a direct pathway of communication between the electrical activity of the brain and an external device. The present paper was aimed to investigate directed connectivity between different areas of the brain during motor imagery (MI)-based BCI. For this purpose, two methods were implemented including, Limited Penetrable Horizontal Visibility Graph (LPHVG) and Direct Lingam. The visibility graph (VG) is a robust algorithm for analyzing complex systems such as the brain. Direct Lingam uses a non-Gaussian model to extract causal links which is appropriate for analyzing large-scale connectivity. First, LPHVG map MI-EEG (electroencephalogram) signals into networks. After extracting the topological features of the networks, a support vector machine classifier was applied to categorize multi-classes MI. The network of all classes was found to be different from one another, and the kappa value of classification was 0.68. The degree sequence of LPHVG was calculated for each channel in order to obtain the direction of brain information flow. Transfer entropy (TE) is used to compute the relations of the channel degree sequence. Therefore, the directed graph between channels was formed. This method is called LPHVG_TE directed graph. The Bayesian network, also known as the Direct LiNGAM model, was implemented for the second method. Finally, images of the LPHVG and Direct Lingam were classified by convolutional neural network (CNN). In this study, Data sets 2a of BCI competition IV was used. The outcomes reveal that the brain network developed by LPHVG (92.7%) might be more effective to distinguish 4 classes of MI than the Direct Lingam (90.6%) and it was shown that graph theory has the potential to get better efficiency of BCI.

Improving cross-subject classification performance of motor imagery signals: a data augmentation-focused deep learning framework

Motor imagery brain-computer interfaces (MI-BCIs) have gained a lot of attention in recent years thanks to their potential to enhance rehabilitation and control of prosthetic devices for individuals with motor disabilities. However, accurate classification of motor imagery signals remains a challenging task due to the high inter-subject variability and non-stationarity in the electroencephalogram (EEG) data. In the context of MI-BCIs, with limited data availability, the acquisition of EEG data can be difficult. In this study, several data augmentation techniques have been compared with the proposed data augmentation technique adaptive cross-subject segment replacement (ACSSR). This technique, in conjunction with the proposed deep learning framework, allows for a combination of similar subject pairs to take advantage of one another and boost the classification performance of MI-BCIs. The proposed framework features a multi-domain feature extractor based on common spatial patterns with a sliding window and a parallel two-branch convolutional neural network. The performance of the proposed methodology has been evaluated on the multi-class BCI Competition IV Dataset 2a through repeated 10-fold cross-validation. Experimental results indicated that the implementation of the ACSSR method (80.47%) in the proposed framework has led to a considerable improvement in the classification performance compared to the classification without data augmentation (77.63%), and other fundamental data augmentation techniques used in the literature. The study contributes to the advancements for the development of effective MI-BCIs by showcasing the ability of the ACSSR method to address the challenges in motor imagery signal classification tasks.

Graph-Based Information Separator and Area Convolutional Network for EEG-Based Intention Decoding

In the current research on brain-computer interface (BCI), the electroencephalography (EEG) signal is usually only represented by a two-dimensional matrix, and the installation position of the EEG electrodes and the correlation between them are not considered. Actually, the cerebral cortex is a continuous potential surface and the information collected directly from each electrode is influenced by the other electrodes, so direct use of the raw data results in information redundancy. This paper converts the EEG signal into a graph, and then creates an information separator (IS) based on the Laplace matrix of the graph to obtain the independent source information of electrode nodes, and propose an IS-based area convolutional network (IS-ACN). Integrating the proposed IS with some advanced methods, the experimental results show that the incorporation of IS can enhance the performance of these methods. By observing and tracking samples with abnormal noise in the BCI competition IV dataset 2a, it is demonstrated that the proposed method can greatly reduce the influence of noise and effectively obtain the source features of EEG signals with low signal-to-noise ratio, and the average accuracy and kappa coefficient of the proposed IS-ACN on this dataset are 80.59% and 74.1%, respectively.

Transfer learning with data alignment and optimal transport for EEG based motor imagery classification

Objective. The non-stationarity of electroencephalogram (EEG) signals and the variability among different subjects present significant challenges in current Brain–Computer Interfaces (BCI) research, which requires a time-consuming specific calibration procedure to address. Transfer Learning (TL) offers a potential solution by leveraging data or models from one or more source domains to facilitate learning in the target domain, so as to address these challenges. Approach. In this paper, a novel Multi-source domain Transfer Learning Fusion (MTLF) framework is proposed to address the calibration problem. Firstly, the method transforms the source domain data with the resting state segment data, in order to decrease the differences between the source domain and the target domain. Subsequently, feature extraction is performed using common spatial pattern. Finally, an improved TL classifier is employed to classify the target samples. Notably, this method does not require the label information of target domain samples, while concurrently reducing the calibration workload. Main results. The proposed MTLF is assessed on Datasets 2a and 2b from the BCI Competition IV. Compared with other algorithms, our method performed relatively the best and achieved mean classification accuracy of 73.69% and 70.83% on Datasets 2a and 2b respectively. Significance. Experimental results demonstrate that the MTLF framework effectively reduces the discrepancy between the source and target domains and acquires better classification performance on two motor imagery datasets.

Selective multi–view time–frequency decomposed spatial feature matrix for motor imagery EEG classification

Decoding brain activity from non-invasive motor imagery electroencephalograph (MI-EEG) has garnered significant attentions for brain-computer interface (BCI) and brain disorders. Notably, owing to the remarkable advances in feature representation, extracting and selecting discriminative features in EEG decoding has gained widespread popularity in recent years. However, many EEG studies suffer from limited sample sizes with low signal-to-noise ratio, making it difficult to effectively represent complementary features in a single view. To address this fundamental limitation, this paper proposes a novel method to Selective extract the Multi-View Time-Frequency decomposed Spatial (S-MVTFS) feature matrix, which employs spatial features from Euclidean and Riemannian spaces based on time–frequency decompositions. It selectively extracts discriminative features on manifold embedded space and classifies feature matrix through sparse support matrix machine. The proposed method has been systematically benchmarked on three BCI competition MI-EEG datasets, and its classification performance surpasses several state-of-the-art methods. Notably, the S-MVTFS method achieved average classification improvements of 0.45%, 2.28%, and 2.04% on BCI-III dataset 4a, BCI-IV dataset 1, and BCI-IV dataset 2a, respectively. Moreover, it effectively captures meaningful temporal varying and spatially coupled features with parameter insensitivity. Our method therefore provides a novel MI-EEG tailored feature representation for decoding brain activity.

EISATC-Fusion: Inception Self-Attention Temporal Convolutional Network Fusion for Motor Imagery EEG Decoding.

The motor imagery brain-computer interface (MI-BCI) based on electroencephalography (EEG) is a widely used human-machine interface paradigm. However, due to the non-stationarity and individual differences among subjects in EEG signals, the decoding accuracy is limited, affecting the application of the MI-BCI. In this paper, we propose the EISATC-Fusion model for MI EEG decoding, consisting of inception block, multi-head self-attention (MSA), temporal convolutional network (TCN), and layer fusion. Specifically, we design a DS Inception block to extract multi-scale frequency band information. And design a new cnnCosMSA module based on CNN and cos attention to solve the attention collapse and improve the interpretability of the model. The TCN module is improved by the depthwise separable convolution to reduces the parameters of the model. The layer fusion consists of feature fusion and decision fusion, fully utilizing the features output by the model and enhances the robustness of the model. We improve the two-stage training strategy for model training. Early stopping is used to prevent model overfitting, and the accuracy and loss of the validation set are used as indicators for early stopping. The proposed model achieves within-subject classification accuracies of 84.57% and 87.58% on BCI Competition IV Datasets 2a and 2b, respectively. And the model achieves cross-subject classification accuracies of 67.42% and 71.23% (by transfer learning) when training the model with two sessions and one session of Dataset 2a, respectively. The interpretability of the model is demonstrated through weight visualization method.

TFTL: A Task-Free Transfer Learning Strategy for EEG-based Cross-Subject & Cross-Dataset Motor Imagery BCI.

Motor imagery-based brain-computer interfaces (MI-BCIs) have been playing an increasingly vital role in neural rehabilitation. However, the long-term task-based calibration required for enhanced model performance leads to an unfriendly user experience, while the inadequacy of EEG data hinders the performance of deep learning models. To address these challenges, a task-free transfer learning strategy (TFTL) for EEG-based cross-subject & cross-dataset MI-BCI is proposed for calibration time reduction and multi-center data co-modeling. TFTL strategy consists of data alignment, shared feature extractor, and specific classifiers, in which the label predictor for MI tasks classification, as well as domain and dataset discriminator for inter-subject variability reduction are concurrently optimized for knowledge transfer from subjects across different datasets to the target subject. Moreover, only resting data of the target subject is used for subject-specific model construction to achieve task-free. We employed three deep learning methods (ShallowConvNet, EEGNet, and TCNet-Fusion) as baseline approaches to evaluate the effectiveness of the proposed strategy on five datasets (BCIC IV Dataset 2a, Dataset 1, Physionet MI, Dreyer 2023, and OpenBMI). The results demonstrate a significant improvement with the inclusion of the TFTL strategy compared to the baseline methods, reaching a maximum enhancement of 15.67% with a statistical significance (p=2.4e-5<0.05). Moreover, task-free resulted in MI trials needed for calibration being 0 for all datasets, which significantly alleviated the calibration burden for patients before usage. The proposed TFTL strategy effectively addresses challenges posed by prolonged calibration periods and insufficient EEG data, thus promoting MI-BCI from laboratory to clinical application.

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.

ADFCNN: Attention-Based Dual-Scale Fusion Convolutional Neural Network for Motor Imagery Brain-Computer Interface.

Convolutional neural networks (CNNs) have been successfully applied to motor imagery (MI)-based brain-computer interface (BCI). Nevertheless, single-scale CNN fail to extract abundant information over a wide spectrum from EEG signals, while typical multi-scale CNNs cannot effectively fuse information from different scales with concatenation-based methods. To overcome these challenges, we propose a new scheme equipped with attention-based dual-scale fusion convolutional neural network (ADFCNN), which jointly extracts and fuses EEG spectral and spatial information at different scales. This scheme also provides novel insight through self-attention for effective information fusion from different scales. Specifically, temporal convolutions with two different kernel sizes identify EEG μ and β rhythms, while spatial convolutions at two different scales generate global and detailed spatial information, respectively, and the self-attention mechanism performs feature fusion based on the internal similarity of the concatenated features extracted by the dual-scale CNN. The proposed scheme achieves the superior performance compared with state-of-the-art methods in subject-specific motor imagery recognition on BCI Competition IV dataset 2a, 2b and OpenBMI dataset, with the cross-session average classification accuracies of 79.39% and significant improvements of 9.14% on BCI-IV2a, 87.81% and 7.66% on BCI-IV2b, 65.26% and 7.2% on OpenBMI dataset, and the within-session average classification accuracies of 86.87% and significant improvements of 10.89% on BCI-IV2a, 87.26% and 8.07% on BCI-IV2b, 84.29% and 5.17% on OpenBMI dataset, respectively. What is more, ablation experiments are conducted to investigate the mechanism and demonstrate the effectiveness of the dual-scale joint temporal-spatial CNN and self-attention modules. Visualization is also used to reveal the learning process and feature distribution of the model.

Improved motor imagery classification using adaptive spatial filters based on particle swarm optimization algorithm.

As a typical self-paced brain-computer interface (BCI) system, the motor imagery (MI) BCI has been widely applied in fields such as robot control, stroke rehabilitation, and assistance for patients with stroke or spinal cord injury. Many studies have focused on the traditional spatial filters obtained through the common spatial pattern (CSP) method. However, the CSP method can only obtain fixed spatial filters for specific input signals. In addition, the CSP method only focuses on the variance difference of two types of electroencephalogram (EEG) signals, so the decoding ability of EEG signals is limited. To make up for these deficiencies, this study introduces a novel spatial filter-solving paradigm named adaptive spatial pattern (ASP), which aims to minimize the energy intra-class matrix and maximize the inter-class matrix of MI-EEG after spatial filtering. The filter bank adaptive and common spatial pattern (FBACSP), our proposed method for MI-EEG decoding, amalgamates ASP spatial filters with CSP features across multiple frequency bands. Through a dual-stage feature selection strategy, it employs the Particle Swarm Optimization algorithm for spatial filter optimization, surpassing traditional CSP approaches in MI classification. To streamline feature sets and enhance recognition efficiency, it first prunes CSP features in each frequency band using mutual information, followed by merging these with ASP features. Comparative experiments are conducted on two public datasets (2a and 2b) from BCI competition IV, which show the outstanding average recognition accuracy of FBACSP. The classification accuracy of the proposed method has reached 74.61 and 81.19% on datasets 2a and 2b, respectively. Compared with the baseline algorithm, filter bank common spatial pattern (FBCSP), the proposed algorithm improves by 11.44 and 7.11% on two datasets, respectively (p < 0.05). It is demonstrated that FBACSP has a strong ability to decode MI-EEG. In addition, the analysis based on mutual information, t-SNE, and Shapley values further proves that ASP features have excellent decoding ability for MI-EEG signals and explains the improvement of classification performance by the introduction of ASP features. These findings may provide useful information to optimize EEG-based BCI systems and further improve the performance of non-invasive BCI.