.png){kind=icon}
.png){kind=icon}
Return
IFBCLNet: Spatio-temporal frequency feature extraction-based MI-EEG classification convolutional network
.png){kind=icon}
PRE
AI
DOI：10.1016/j.bspc.2024.106092.png)
.png){kind=icon}
Original request for help.png){kind=icon}
Share.png){kind=icon}
Save
Abstract
En 中文
Brain-computer interfaces (BCIs) provide a way for individuals to interact with and control external devices using their brain signals. Among the most promising BCI methods, the motor imagery (MI) paradigm-based electroencephalogram (EEG) is widely used due to its non-invasive instrumentation and ease of experimentation. However, traditional decoding methods face challenges in extracting spatio-temporal-frequency features from EEG signals. In this paper, we present a novel network framework (IFBCLNet) that integrates an interpretable filter bank with the convolutional neural network (CNN) and the long and short-term memory (LSTM) module. IFBCLNet demonstrates unique spatio-temporal frequency feature extraction capabilities, allowing for accurate interpretation of human intentions. Extensive experiments conducted on three EEG datasets (BCICIV-2a, BCICIV-2b, and High Gamma Dataset) reveal that our proposed framework achieves high accuracy rates of 78.79%, 87.76%, and 95.35%, respectively. The results also show the superiority of our model over recent baseline models. Additionally, the cross-subject experiments on the BCICIV-2a, BCICIV2b, and HGD datasets achieved high accuracies of 81.23%, 90.29%, and 96.74%, respectively, indicate that our model can well deal with different types of intentions from various subjects, which is more suitable for BCI-based applications.
Keywords:
Brain-computer interface (BCI)
Motor imagery EEG (MI-EEG) decoding
Interpretable filter bank
Convolutional neural network (CNN)
Long and short-term memory (LSTM)
Journal
IF:
4.9
Papers:
9.8K
Citations:
2.4W