MI-DABAN: A dual-attention-based adversarial network for motor imagery classification

Li, Huiying; Zhang, Dongxue; Xie, Jingmeng

doi:10.1016/j.compbiomed.2022.106420

Cited by 27 publications

(9 citation statements)

References 45 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Year Acc Kappa FBCSP [13] 2008 67.75% 0.5700 CCSP [14] 2009 66.51% 0.5535 ConvNet [16] 2017 72.53% 0.6337 EEGNet [17] 2018 74.61% 0.6615 Incep-EEGNet [18] 2020 74.07% 0.6543 TS-SEFFNet [26] 2021 74.71% 0.6628 MRGF [46] 2022 70.11% 0.6015 MI-DABAN [27] 2023 76.16% 0.6821 SCPGE [47] 2023 68.64% 0.5817 Our Method 2024 77.89% 0.7052 Ang et al [13] decomposed the original EEG signals into multiple frequency bands using filter banks and then extracted the features using the common spatial pattern (CSP) method, achieving an accuracy of 67.75% and a Kappa value of 0.57 on the BCI-IV-2a database. Kang et al [14] considered the relationship between the covariance matrices of different subjects and obtained a composite covariance matrix by weighted averaging the covariance matrices of subjects in the dataset.…”

Section: Methodsmentioning

confidence: 99%

“…In this study, data from the 3 EOG electrodes were removed to prevent interference from eye movement artifacts. Performing motor imagery tasks induces task-related synchronization (ERD) and task-related desynchronization (ERS) in different brain regions, with the main perceptual motor rhythms being mu (8-13 Hz) and beta waves (18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29)(30). The specific frequency ranges vary from person to person, thus bandpass filtering from 4-40 Hz was applied to the raw data to minimize the negative effects while retaining the MI task-related features.…”

Section: Preprocessing and Experimental Setupmentioning

confidence: 99%

“…Therefore, the network itself needs to have strong feature extraction capabilities. Some studies [26,27] have employed attention mechanisms to extract sufficiently discriminative features, but they have not addressed the issue of the small data volume. Other studies [28,29] have attempted to expand the original dataset using data augmentation methods, yielding promising results.…”

Section: Introductionmentioning

confidence: 99%

See 2 more Smart Citations

MST-DGCN: A Multi-Scale Spatio-Temporal and Dynamic Graph Convolution Fusion Network for Electroencephalogram Recognition of Motor Imagery

Chen,

Liu,

2024

Electronics

View full text Add to dashboard Cite

The motor imagery brain-computer interface (MI-BCI) has the ability to use electroencephalogram (EEG) signals to control and communicate with external devices. By leveraging the unique characteristics of task-related brain signals, this system facilitates enhanced communication with these devices. Such capabilities hold significant potential for advancing rehabilitation and the development of assistive technologies. In recent years, deep learning has received considerable attention in the MI-BCI field due to its powerful feature extraction and classification capabilities. However, two factors significantly impact the performance of deep-learning models. The size of the EEG datasets influences how effectively these models can learn. Similarly, the ability of classification models to extract features directly affects their accuracy in recognizing patterns. In this paper, we propose a Multi-Scale Spatio-Temporal and Dynamic Graph Convolution Fusion Network (MST-DGCN) to address these issues. In the data-preprocessing stage, we employ two strategies, data augmentation and transfer learning, to alleviate the problem of an insufficient data volume in deep learning. By using multi-scale convolution, spatial attention mechanisms, and dynamic graph neural networks, our model effectively extracts discriminative features. The MST-DGCN mainly consists of three parts: the multi-scale spatio-temporal module, which extracts multi-scale information and refines spatial attention; the dynamic graph convolution module, which extracts key connectivity information; and the classification module. We conduct experiments on real EEG datasets and achieve an accuracy of 77.89% and a Kappa value of 0.7052, demonstrating the effectiveness of the MST-DGCN in MI-BCI tasks. Our research provides new ideas and methods for the further development of MI-BCI systems.

show abstract

Section: Methodsmentioning

confidence: 99%

Section: Preprocessing and Experimental Setupmentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

MST-DGCN: A Multi-Scale Spatio-Temporal and Dynamic Graph Convolution Fusion Network for Electroencephalogram Recognition of Motor Imagery

Chen,

Liu,

2024

Electronics

View full text Add to dashboard Cite

show abstract

“… Li et al (2023) proposed a new dual-attention-based MI classification adversarial network MI-DABAN. This network can reduce the distributional differences between domains by analyzing the output differences between two classifiers and can increase the distance between the samples of confusing target domains and the decision boundary to improve the classification performance ( Li et al, 2023 ). Milanés Hermosilla et al (2021) used the Shallow Convolutional Network to classify and recognize MI-EEG signals with excellent results ( Milanés Hermosilla et al, 2021 ).…”

Section: Introductionmentioning

confidence: 99%

EMPT: a sparsity Transformer for EEG-based motor imagery recognition

Liu,

Shi

et al. 2024

Front. Neurosci.

View full text Add to dashboard Cite

IntroductionTransformer network is widely emphasized and studied relying on its excellent performance. The self-attention mechanism finds a good solution for feature coding among multiple channels of electroencephalography (EEG) signals. However, using the self-attention mechanism to construct models on EEG data suffers from the problem of the large amount of data required and the complexity of the algorithm.MethodsWe propose a Transformer neural network combined with the addition of Mixture of Experts (MoE) layer and ProbSparse Self-attention mechanism for decoding the time-frequency-spatial domain features from motor imagery (MI) EEG of spinal cord injury patients. The model is named as EEG MoE-Prob-Transformer (EMPT). The common spatial pattern and the modified s-transform method are employed for achieving the time-frequency-spatial features, which are used as feature embeddings to input the improved transformer neural network for feature reconstruction, and then rely on the expert model in the MoE layer for sparsity mapping, and finally output the results through the fully connected layer.ResultsEMPT achieves an accuracy of 95.24% on the MI EEG dataset for patients with spinal cord injury. EMPT has also achieved excellent results in comparative experiments with other state-of-the-art methods.DiscussionThe MoE layer and ProbSparse Self-attention inside the EMPT are subjected to visualisation experiments. The experiments prove that sparsity can be introduced to the Transformer neural network by introducing MoE and kullback-leibler divergence attention pooling mechanism, thereby enhancing its applicability on EEG datasets. A novel deep learning approach is presented for decoding EEG data based on MI.

show abstract

“…They converted the EEGs into time series segments and then calculated the connectivity features between EEG electrodes in every segment via 2D CNN and finally fed the feature vectors to the LSTM network for training. Li et al [23] proposed a new dual-attention-based adversarial network for motor imagery classification. Their framework uses multi-subject knowledge to enhance the classification performance of single-subject motor imagery tasks through intelligently utilizing a new adversarial learning algorithm and two unshared attention blocks.…”

Section: Introductionmentioning

confidence: 99%

Automated Motor Imagery Detection Through EEG Analysis and Deep Learning Models for Brain-Computer Interface Applications

Li,

Liu,

Tian

2024

ijacsa

View full text Add to dashboard Cite

The classification of motor imagery holds significant importance within brain-computer interface (BCI) research as it allows for the identification of a person's intention, such as controlling a prosthesis. Motor imagery involves the brain's dynamic activities, commonly captured using electroencephalography (EEG) to record nonstationary time series with low signal-to-noise ratios. While various methods exist for extracting features from EEG signals, the application of deep learning techniques to enhance the representation of EEG features for improved motor imagery classification performance has been relatively unexplored. This research introduces a new deep learning approach based on two-dimensional CNNs with different architectures. Specifically, time-frequency domain representations of EEGs obtained by the wavelet transform method with different mother wavelets (Mexicanhat, Cmor, and Cgaus). The BCI competition IV-2a dataset held in 2008 was utilized for testing the proposed deep learning approaches. Several experiments were conducted and the results showed that the proposed method achieved better performance than some state-ofthe-art methods. The findings of this study showed that the architecture of CNN and specifically the number of convolution layers in this deep learning network has a significant effect on the classification performance of motor imagery brain data. In addition, the mother wavelet in the wavelet transform is very important in the classification performance of motor imagery EEG data.

show abstract

MI-DABAN: A dual-attention-based adversarial network for motor imagery classification

Cited by 27 publications

References 45 publications

MST-DGCN: A Multi-Scale Spatio-Temporal and Dynamic Graph Convolution Fusion Network for Electroencephalogram Recognition of Motor Imagery

MST-DGCN: A Multi-Scale Spatio-Temporal and Dynamic Graph Convolution Fusion Network for Electroencephalogram Recognition of Motor Imagery

EMPT: a sparsity Transformer for EEG-based motor imagery recognition

Automated Motor Imagery Detection Through EEG Analysis and Deep Learning Models for Brain-Computer Interface Applications

Contact Info

Product

Resources

About