Cauchy non-convex sparse feature selection method for the high-dimensional small-sample problem in motor imagery EEG decoding

Frontiers in Neuroscience, Год журнала: 2023, Номер 17

Опубликована: Ноя. 3, 2023

The time, frequency, and space information of electroencephalogram (EEG) signals is crucial for motor imagery decoding. However, these temporal-frequency-spatial features are high-dimensional small-sample data, which poses significant challenges Sparse regularization an effective method addressing this issue. the most commonly employed sparse models in decoding, such as least absolute shrinkage selection operator (LASSO), a biased estimation leads to loss target feature information. In paper, we propose non-convex model that employs Cauchy function. By designing proximal gradient algorithm, our proposed achieves closer-to-unbiased than existing models. Therefore, it can learn more accurate, discriminative, Additionally, perform classification simultaneously, without requiring additional classifiers. We conducted experiments on two publicly available EEG datasets. achieved average accuracy 82.98% 64.45% subject-dependent subject-independent decoding assessment methods, respectively. experimental results show significantly improve performance with better deep learning methods. Furthermore, shows generalization capability, parameter consistency over different datasets robust across training sample sizes. Compared converges faster, shorter time.

Язык: Английский

---

Deep CNN‐based classification of motor imagery tasks from EEG signals using 2D wavelet transformed images of adaptively reconstructed signals from MVMD decomposed modes

International Journal of Imaging Systems and Technology, Год журнала: 2023, Номер 33(6), С. 1988 - 2011

Опубликована: Май 22, 2023

Abstract Brain–computer interfaces (BCI) have begun to revolutionize many aspects of human life, ranging from health smart living and communication devices. Therefore, BCI technologies require accurate recognition classification systems the brain responses a variety motor imagery (MI) movements through electroencephalogram (EEG) signals. Nowadays, deep learning models received lot interest for features classifying numerous kinds data. However, it has not been thoroughly investigated EEG signal classification. In this study, reach an improved performance in MI‐EEG classification, we proposed new strategy based on continuous wavelet transform (CWT)‐based time‐frequency maps generate two‐dimensional (2D) images adaptively reconstructed signals extracted multivariate mode decomposition (MVMD) modes. next step, are projected space using multiclass common spatial pattern (MCCSP) filtering. The resulting then fed convolutional neural network (CNN) architectures (AlexNet LeNet). framework benefit achieving high accuracy be attained even with significant amount input LeNet AlexNet best average rates 95.33% 93.66% dataset 1 competition IV results 2a more promising than current state arts. Our depict that task image approaches, along CNNs, is as comparable or superior other existing traditional approaches provides potential upcoming research.

Язык: Английский

---

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

Опубликована: Авг. 23, 2023

<p>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. The block extracts multi-scale features, MSA increases global time-dependence then TCN high-level features. fusion consists feature decision fusion, fully utilizing features output by enhances robustness model. We improve two-stage training strategy training. Early stopping is used to prevent overfitting, accuracy loss validation set are as indicators early stopping. proposed achieves within-subject classification accuracies 83.18\% 87.44\% on BCI Competition IV Datasets 2a 2b, respectively. And cross-subject 65.37\% 65.62\% (by transfer learning) when with two sessions one session Dataset 2a, code can be obtained at https://github.com/LiangXiaohan506/EISATC-Fusion.</p>

Язык: Английский

---

Research on MI EEG signal classification algorithm using multi-model fusion strategy coupling

Computer Methods in Biomechanics & Biomedical Engineering, Год журнала: 2023, Номер 28(1), С. 51 - 60

Опубликована: Ноя. 20, 2023

To enhance the accuracy of motor imagery(MI)EEG signal recognition, two methods, namely power spectral density and wavelet packet decomposition combined with a common spatial pattern, were employed to explore feature information in depth MI EEG signals. The extracted features subjected series fusion, F-test method was used select higher content. Here regarding classification, we further proposed Platt Scaling probability calibration calibrate results obtained from six basic classifiers, random forest (RF), support vector machines (SVM), Logistic Regression (LR), Gaussian naïve bayes (GNB), eXtreme Gradient Boosting (XGBoost), Light Machine (LightGBM). From these 12 three four selected for model fusion. validated on Datasets 2a 4th International BCI Competition, achieving an average data nine subjects reached 91.46%, which indicates that fusion effective improve classification accuracy, provides some reference value research brain-machine interface.

Язык: Английский

---

MBCNN-EATCFNet: A multi-branch neural network with efficient attention mechanism for decoding EEG-based motor imagery

Robotics and Autonomous Systems, Год журнала: 2024, Номер unknown, С. 104899 - 104899

Опубликована: Дек. 1, 2024

Язык: Английский

---