A Multi-Branch Convolutional Neural Network with Squeeze-and-Excitation Attention Blocks for EEG-Based Motor Imagery Signals Classification

Diagnostics, Journal Year: 2022, Volume and Issue: 12(4), P. 995 - 995

Published: April 15, 2022

Electroencephalography-based motor imagery (EEG-MI) classification is a critical component of the brain-computer interface (BCI), which enables people with physical limitations to communicate outside world via assistive technology. Regrettably, EEG decoding challenging because complexity, dynamic nature, and low signal-to-noise ratio signal. Developing an end-to-end architecture capable correctly extracting data's high-level features remains difficulty. This study introduces new model for MI known as Multi-Branch EEGNet squeeze-and-excitation blocks (MBEEGSE). By clearly specifying channel interdependencies, multi-branch CNN attention employed adaptively change channel-wise feature responses. When compared existing state-of-the-art models, suggested achieves good accuracy (82.87%) reduced parameters in BCI-IV2a dataset (96.15%) high gamma dataset.

Language: Английский

---

State-of-the-Art on Brain-Computer Interface Technology

Sensors, Journal Year: 2023, Volume and Issue: 23(13), P. 6001 - 6001

Published: June 28, 2023

This paper provides a comprehensive overview of the state-of-the-art in brain–computer interfaces (BCI). It begins by providing an introduction to BCIs, describing their main operation principles and most widely used platforms. The then examines various components BCI system, such as hardware, software, signal processing algorithms. Finally, it looks at current trends research related use for medical, educational, other purposes, well potential future applications this technology. concludes highlighting some key challenges that still need be addressed before widespread adoption can occur. By presenting up-to-date assessment technology, will provide valuable insight into where field is heading terms progress innovation.

Language: Английский

---

A Cross-Space CNN With Customized Characteristics for Motor Imagery EEG Classification

IEEE Transactions on Neural Systems and Rehabilitation Engineering, Journal Year: 2023, Volume and Issue: 31, P. 1554 - 1565

Published: Jan. 1, 2023

The classification of motor imagery-electroen-cephalogram( MI-EEG)based brain-computer interface(BCI) can be used to decode neurological activities, which has been widely applied in the control external devices. However, two factors still hinder improvement accuracy and robustness, especially multi-class tasks. First, existing algorithms are based on a single space (measuring or source space). They suffer from holistic low spatial resolution measuring locally high information accessed space, failing provide high-resolution representations. Second, subject specificity is not sufficiently characterized, resulting loss personalized intrinsic information. Therefore, we propose cross-space convolutional neural network (CS-CNN) with customized characteristics for four-class MI-EEG classification. This algorithm uses modified band common patterns (CBCSP) duplex mean-shift clustering (DMSClustering) express specific rhythms distribution cross-space. At same time, multi-view features frequency domains extracted, connecting CNN fuse spaces classify them. was collected 20 subjects. Lastly, proposed 96.05% real MRI 94.79% without private dataset. And results BCI competition Ⅳ-2a show that CS-CNN outperforms state-of-the-art algorithms, achieving an 1.98%, standard deviation reduction 5.15%.

Language: Английский

---

A novel multi-branch hybrid neural network for motor imagery EEG signal classification

Biomedical Signal Processing and Control, Journal Year: 2022, Volume and Issue: 77, P. 103718 - 103718

Published: April 28, 2022

Language: Английский

---

Automated Detection of Mechanical Damage in Flaxseeds Using Radiographic Imaging and Machine Learning

Food and Bioprocess Technology, Journal Year: 2022, Volume and Issue: 16(3), P. 526 - 536

Published: Nov. 28, 2022

Language: Английский

---

Human Fall Detection Using 3D Multi-Stream Convolutional Neural Networks with Fusion

Diagnostics, Journal Year: 2022, Volume and Issue: 12(12), P. 3060 - 3060

Published: Dec. 6, 2022

Human falls, especially for elderly people, can cause serious injuries that might lead to permanent disability. Approximately 20-30% of the aged people in United States who experienced fall accidents suffer from head trauma, injuries, or bruises. Fall detection is becoming an important public healthcare problem. Timely and accurate incident could enable instant delivery medical services injured. New advances vision-based technologies, including deep learning, have shown significant results action recognition, where some focus on actions. In this paper, we propose automatic human system using multi-stream convolutional neural networks with fusion. The based a multi-level image-fusion approach every 16 frames input video highlight movement differences within range. This four consecutive preprocessed images are fed new proposed efficient lightweight CNN model four-branch architecture (4S-3DCNN) classifies whether there fall. evaluation included use more than 6392 generated sequences Le2i dataset, which publicly available dataset. method, three-fold cross-validation validate generalization susceptibility overfitting, achieved 99.03%, 99.00%, 99.68%, 99.00% accuracy, sensitivity, specificity, precision, respectively. experimental prove outperforms state-of-the-art models, GoogleNet, SqueezeNet, ResNet18, DarkNet19, detection.

Language: Английский

---

Electroencephalogram-Based Motor Imagery Signals Classification Using a Multi-Branch Convolutional Neural Network Model with Attention Blocks

Bioengineering, Journal Year: 2022, Volume and Issue: 9(7), P. 323 - 323

Published: July 18, 2022

Brain signals can be captured via electroencephalogram (EEG) and used in various brain-computer interface (BCI) applications. Classifying motor imagery (MI) using EEG is one of the important applications that help a stroke patient to rehabilitate or perform certain tasks. Dealing with EEG-MI challenging because are weak, may contain artefacts, dependent on patient's mood posture, have low signal-to-noise ratio. This paper proposes multi-branch convolutional neural network model called Multi-Branch EEGNet Convolutional Block Attention Module (MBEEGCBAM) attention mechanism fusion techniques classify signals. The applied both channel-wise spatial-wise. proposed lightweight has fewer parameters higher accuracy compared other state-of-the-art models. 82.85% 95.45% BCI-IV2a dataset high gamma dataset, respectively. Additionally, when approach (FMBEEGCBAM), it achieves 83.68% 95.74% accuracy,

Language: Английский

---

Emotion Recognition via Multiscale Feature Fusion Network and Attention Mechanism

IEEE Sensors Journal, Journal Year: 2023, Volume and Issue: 23(10), P. 10790 - 10800

Published: April 13, 2023

Traditional manual feature-based machine learning methods and deep networks have been used for electroencephalogram (EEG)-based emotion recognition in recent years. However, some existing studies ignore the low signal-to-noise ratio fact that each subject has unique EEG traits, which suffer from accuracy poor robustness. To solve these problems, we propose a novel attention mechanism-based multiscale feature fusion network (AM-MSFFN) considers high-level features at different scales to improve generalization of model subjects. Specifically, first utilize spatial-temporal convolutional block extract temporal spatial signals sequentially. Subsequently, considering sampling rate signals, separable convolutions are designed capturing emotional state-related information, better combine output mapping relationships. Convolutional module mechanism (CBAM) is applied after point-wise convolution, handle variations subjects key information facilitates classification. In addition, adopt preprocessing based on data augmentation alignment quality training samples. Moreover, ablation show proposed convolution contribute significant consistent gain performance our AM-MSFFN model. verify effectiveness algorithm, conducted extensive experiments DEAP dataset SEED. The average accuracies achieve 99.479% 99.297% arousal valence, respectively. results demonstrated feasibility method.

Language: Английский

---

A robust multi-branch multi-attention-mechanism EEGNet for motor imagery BCI decoding

Journal of Neuroscience Methods, Journal Year: 2024, Volume and Issue: 405, P. 110108 - 110108

Published: March 6, 2024

Language: Английский

---

An in-depth survey on Deep Learning-based Motor Imagery Electroencephalogram (EEG) classification

Artificial Intelligence in Medicine, Journal Year: 2023, Volume and Issue: 147, P. 102738 - 102738

Published: Dec. 2, 2023

Electroencephalogram (EEG)-based Brain-Computer Interfaces (BCIs) build a communication path between human brain and external devices. Among EEG-based BCI paradigms, the most commonly used one is motor imagery (MI). As hot research topic, MI has largely contributed to medical fields smart home industry. However, because of low signal-to-noise ratio (SNR) non-stationary characteristic EEG data, it difficult correctly classify different types MI-EEG signals. Recently, advances in Deep Learning (DL) significantly facilitate development BCIs. In this paper, we provide systematic survey DL-based classification methods. Specifically, first comprehensively discuss several important aspects classification, covering input formulations, network architectures, public datasets, etc. Then, summarize problems model performance comparison give guidelines future studies for fair comparison. Next, fairly evaluate representative models using source code released by authors meticulously analyse evaluation results. By performing ablation study on architecture, found that (1) effective feature fusion indispensable multi-stream CNN-based models. (2) LSTM should be combined with spatial extraction techniques obtain good performance. (3) use dropout contributes little improving performance, (4) adding fully connected layers increases their parameters but might not improve Finally, raise open issues possible directions.

Language: Английский

---