Article
Engineering, Biomedical
Jingfeng Bi, Ming Chu
Summary: The goal of this study is to design a single-limb, multi-category motor imagery paradigm and achieve cross-subject intention recognition through the transfer data learning network (TDLNet). The network processes cross-subject EEG signals and assigns weights to signal channels using the Residual Attention Mechanism Module (RAMM), resulting in the best classification results.
IEEE TRANSACTIONS ON NEURAL SYSTEMS AND REHABILITATION ENGINEERING
(2023)
Article
Biology
Guoyang Liu, Lan Tian, Weidong Zhou
Summary: In this paper, a method for optimizing the performance and computational complexity of multiclass MI-BCI is proposed. It utilizes multiscale time-frequency segmentation and feature selection to extract discriminative features, and utilizes SVM for class prediction. Experimental results show that the method achieves high accuracy on multiple datasets and enables real-time prediction.
COMPUTERS IN BIOLOGY AND MEDICINE
(2022)
Article
Computer Science, Interdisciplinary Applications
Fangzhou Xu, Xiaoyan Xu, Yanan Sun, Jincheng Li, Gege Dong, Yuandong Wang, Han Li, Lei Wang, Yingchun Zhang, Shaopeng Pang, Sen Yin
Summary: This study introduces a long short-term memory recurrent neural network for decoding electroencephalogram or electrocorticogram and implementing an effective brain-computer interface system. By combining the decoded features with a gradient boosting classifier, high recognition accuracies can be achieved.
COMPUTER METHODS AND PROGRAMS IN BIOMEDICINE
(2022)
Review
Computer Science, Artificial Intelligence
Dongrui Wu, Xue Jiang, Ruimin Peng
Summary: A brain-computer interface (BCI) allows users to communicate with external devices using brain signals, and transfer learning (TL) has been widely used in MI-based BCIs to reduce calibration effort and improve utility.
Article
Chemistry, Analytical
Ilaria Siviero, Gloria Menegaz, Silvia Francesca Storti
Summary: In this study, a new signal processing framework is proposed to address the challenge of obtaining discriminative features in motor-imagery brain-computer interfaces (MI-BCIs). The framework combines translation-invariant features (TIFs) with functional connectivity features (BCFs) using a feature fusion approach and achieves better performance compared to existing methods. The results highlight the importance of merging TIFs with BCFs for improved classification in MI-BCIs.
Article
Multidisciplinary Sciences
Rabia Avais Khan, Nasir Rashid, Muhammad Shahzaib, Umar Farooq Malik, Arshia Arif, Javaid Iqbal, Mubasher Saleem, Umar Shahbaz Khan, Mohsin Tiwana
Summary: Robotics and artificial intelligence have played a significant role in developing assistive technologies for people with motor disabilities. In this study, a novel framework for classifying binary-class electroencephalogram (EEG) data has been proposed. The framework achieved high classification accuracies with logistic regression classifier on two datasets, indicating its potential for real-time Brain-Computer Interface (BCI) systems and 2-class Motor Imagery signals classification applications.
Article
Chemistry, Analytical
Zhanyuan Chang, Congcong Zhang, Chuanjiang Li
Summary: The study focused on the application of attention mechanism-based multi-scale convolution network and transfer learning in brain-computer interface systems, and improved the classification recognition rate of EEG signals through comparison of different algorithms.
Article
Engineering, Biomedical
Minmin Miao, Zhong Yang, Hong Zeng, Wenbin Zhang, Baoguo Xu, Wenjun Hu
Summary: This paper proposes a novel explainable cross-task adaptive transfer learning method for decoding motor imagery electroencephalography (MI EEG). The method achieves effective decoding model by pre-training with extensive motor execution EEG data and fine-tuning with partial MI EEG data. Experimental results demonstrate its superiority over state-of-the-art algorithms and the effectiveness of cross-task adaptation is further validated through explainability analysis.
JOURNAL OF NEURAL ENGINEERING
(2023)
Article
Engineering, Biomedical
Siwei Liu, Jia Zhang, Andong Wang, Hanrui Wu, Qibin Zhao, Jinyi Long
Summary: This paper introduces a novel end-to-end deep subject adaptation convolutional neural network (SACNN) for EEG-based MI classification. By extracting both temporal and spatial features and reducing feature distribution shift, SACNN achieves significant improvement in classification accuracy.
JOURNAL OF NEURAL ENGINEERING
(2022)
Article
Biology
Xiao-Cong Zhong, Qisong Wang, Dan Liu, Jing-Xiao Liao, Runze Yang, Sanhe Duan, Guohua Ding, Jinwei Sun
Summary: In this paper, a deep domain adaptation framework with correlation alignment (DDAF-CORAL) is proposed to address the problem of distribution divergence in motor imagery classification across domains.
COMPUTERS IN BIOLOGY AND MEDICINE
(2023)
Article
Computer Science, Artificial Intelligence
Xin Zhang, Zhengqing Miao, Carlo Menon, Yelong Zheng, Meirong Zhao, Dong Ming
Summary: This paper proposes a Siamese deep domain adaptation (SDDA) framework for cross-session MI classification, which preprocesses and aligns EEG signals using mathematical models to improve classification accuracy. The framework outperforms state-of-the-art transfer learning methods.
Article
Chemistry, Analytical
Xuying Wang, Rui Yang, Mengjie Huang
Summary: This study proposes an unsupervised deep-transfer-learning-based method to address the limitations of brain-computer interface (BCI) systems. By aligning data in Euclidean space and extracting features using common spatial pattern (CSP), the method achieves EEG signal classification through deep convolutional neural network (CNN). Experimental results demonstrate the effectiveness of the proposed method.
Article
Engineering, Biomedical
Xiuyu Huang, Shuang Liang, Yuanpeng Zhang, Nan Zhou, Witold Pedrycz, Kup-Sze Choi
Summary: This paper proposes a few-shot learning method called temporal episode relation learning (TERL) for generating a reliable model for a target subject with limited MI trials in practical BCI applications. TERL compares MI trials through episode-based training and can be directly applied to new users, improving user experience and enabling real-world MIBCI applications.
IEEE TRANSACTIONS ON NEURAL SYSTEMS AND REHABILITATION ENGINEERING
(2023)
Article
Engineering, Biomedical
Do-Yeun Lee, Ji-Hoon Jeong, Byeong-Hoo Lee, Seong-Whan Lee
Summary: The study focused on decoding various forearm movements from EEG signals using a small number of samples. A convolutional neural network based on inter-task transfer learning was proposed, achieving improved classification performance by training the reconstructed ME-EEG signals together with a small amount of MI-EEG signals. The proposed method showed increased performance compared to conventional models, suggesting the feasibility of BCI learning strategies with stable performance using a small calibration dataset and time.
IEEE TRANSACTIONS ON NEURAL SYSTEMS AND REHABILITATION ENGINEERING
(2022)
Article
Computer Science, Artificial Intelligence
Donglin Li, Jianhui Wang, Jiacan Xu, Xiaoke Fang, Ying Ji
Summary: A cross-channel specific-mutual feature transfer learning (CCSM-FT) network model is proposed in this paper to address the issue of extracting specific and mutual features from multiregion signals in the brain. Effective training tricks are used to maximize the distinction between these two types of features and improve algorithm effectiveness.
IEEE TRANSACTIONS ON NEURAL NETWORKS AND LEARNING SYSTEMS
(2023)