A multi-branch, multi-scale, and multi-view CNN with lightweight temporal attention mechanism for EEG-based motor imagery decoding

Computer Methods in Biomechanics & Biomedical Engineering, Journal Year: 2025, Volume and Issue: unknown, P. 1 - 15

Published: Jan. 6, 2025

Convolutional neural networks (CNNs) have been widely utilized for decoding motor imagery (MI) from electroencephalogram (EEG) signals. However, extracting discriminative spatial-temporal-spectral features low signal-to-noise ratio EEG signals remains challenging. This paper proposes MBMSNet , a multi-branch, multi-scale, and multi-view CNN with lightweight temporal attention mechanism EEG-Based MI decoding. Specifically, first extracts representations raw signals, followed by independent branches to capture spatial, spectral, temporal-spatial, temporal-spectral features. Each branch includes domain-specific convolutional layer, variance layer. Finally, the derived each are concatenated weights classified through fully connected Experiments demonstrate outperforms state-of-the-art models, achieving accuracies of 84.60% on BCI Competition IV 2a, 87.80% 2b, 74.58% OpenBMI, showcasing its potential robust applications.

Language: Английский

---

An improved multi-scale convolution and transformer network for EEG-based motor imagery decoding

International Journal of Machine Learning and Cybernetics, Journal Year: 2025, Volume and Issue: unknown

Published: Feb. 4, 2025

Language: Английский

---

MS-HyFS: A novel multiscale hybrid framework with Scalable electrodes for motor imagery classification

Biomedical Signal Processing and Control, Journal Year: 2025, Volume and Issue: 106, P. 107706 - 107706

Published: Feb. 21, 2025

Language: Английский

---

Study on real-time warning system of blind path for the visually impaired based on improved deep residual shrinkage network

Scientific Reports, Journal Year: 2025, Volume and Issue: 15(1)

Published: April 29, 2025

Visually impaired individuals often face various obstacles when navigating blind roads, such as road disconnections, obstructions, and more complex emergencies, which can leave them in difficult situations. Traditional early warning methods suffer from low accuracy lack real-time capabilities. Therefore, this study proposes a novel system for traffic jams on roads. By analyzing the emotional state (normal, mild anxiety, extreme anxiety) electroencephalogram (EEG) signals of visually they are trapped, determine whether distress require assistance. Additionally, considering complexity environment fact that EEG prone to external interference during acquisition, introduces an improved deep residual shrinkage network based dense blocks (DB-DRSN). DB-DRSN replaces convolutional hidden layer original module with integrates connections optimize use both shallow features. The results show achieves 96.72% recognizing difficulties faced by impaired, significantly outperforming traditional models. Compared other methods, proposed offers quicker assistance individuals. demonstrated strong performance detecting about jams, greatly enhancing safety enabling timely detection intervention.

Language: Английский

---

CCLNet: multiclass motor imagery EEG decoding through extended common spatial patterns and CNN-LSTM hybrid network

The Journal of Supercomputing, Journal Year: 2025, Volume and Issue: 81(7)

Published: May 1, 2025

Language: Английский

---

A novel approach for communicating with patients suffering from completely locked-in-syndrome (CLIS) via thoughts: Brain computer interface system using EEG signals and artificial intelligence

Neuroscience Informatics, Journal Year: 2023, Volume and Issue: 3(2), P. 100126 - 100126

Published: March 21, 2023

This paper investigates the development of an intelligent system method to address completely locked-in-syndrome (CLIS) that is caused by some illnesses such as Amyotrophic Lateral Sclerosis (ALS) most predominant type Motor Neuron Disease (MND). In last stages ALS and despite limitations in body movements, patients however will have a fully functional brain cognitive capabilities able feel pain but fail communicate. aims CLIS problem utilizing EEG signals human generates when thinking about specific feeling or imagination way The aim develop low-cost affordable for use communicate with carers family members. this paper, novel implementation ASPS (Automated Sensor Signal Processing Selection) approach feature extraction presented select suitable Sensory Characteristic Features (SCFs) detect thoughts imaginations. Artificial Neural Networks (ANN) are used verify results. findings show capture information can be means communication; allows selection important features reliable communication. explains validation signal classification bespoke arrangement. Hence, future work present results relatively high number volunteers, sensors processing methods.

Language: Английский

---

Motor Imagery Decoding Using Ensemble Curriculum Learning and Collaborative Training

Published: Feb. 26, 2024

In this work, we study the problem of cross-subject motor imagery (MI) decoding from electroencephalography (EEG) data. Multi-subject EEG datasets present several kinds domain shifts due to various inter-individual differences (e.g. brain anatomy, personality and cognitive profile). These render multi-subject training a challenging task also impede robust generalization. Inspired by importance generalization techniques for tackling such issues, propose two-stage model ensemble architecture built with multiple feature extractors (first stage) shared classifier (second stage), which train end-toend two novel loss terms. The first applies curriculum learning, forcing each extractor specialize subset subjects promoting diversity. second is an intra-ensemble distillation objective that allows collaborative exchange knowledge between models ensemble. We compare our method against state-of-the-art techniques, conducting subject-independent experiments on large MI datasets, namely PhysioNet OpenBMI. Our algorithm outperforms all methods in both 5-fold cross-validation leave-one-subject-out evaluation settings, using substantially lower number trainable parameters. demonstrate ensembling approach combining powers learning training, leads high capacity performance. work addresses issue paving way calibration-free brain-computer interfaces. make code publicly available at: https://github.com/gzoumpourlis/Ensemble-MI.

Language: Английский

---

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

Journal of Neural Engineering, Journal Year: 2023, Volume and Issue: 20(4), P. 046031 - 046031

Published: Aug. 1, 2023

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

Language: Английский

---

MI-MBFT: Superior Motor Imagery Decoding of Raw EEG Data Based on a Multi-Branch and Fusion Transformer Framework

IEEE Sensors Journal, Journal Year: 2024, Volume and Issue: 24(21), P. 34879 - 34891

Published: Aug. 28, 2024

Language: Английский

---

Enhancing motor imagery classification: a novel CNN with self-attention using local and global features of filtered EEG data

Connection Science, Journal Year: 2024, Volume and Issue: 36(1)

Published: Nov. 16, 2024

Language: Английский

---