EEGformer: A transformer–based brain activity classification method using EEG signal

Abstract

BackgroundThe effective analysis methods for steady-state visual evoked potential (SSVEP) signals are critical in supporting an early diagnosis of glaucoma. Most efforts focused on adopting existing techniques to the SSVEPs-based brain–computer interface (BCI) task rather than proposing new ones specifically suited to the domain.MethodGiven that electroencephalogram (EEG) signals possess temporal, regional, and synchronous characteristics of brain activity, we proposed a transformer–based EEG analysis model known as EEGformer to capture the EEG characteristics in a unified manner. We adopted a one-dimensional convolution neural network (1DCNN) to automatically extract EEG-channel-wise features. The output was fed into the EEGformer, which is sequentially constructed using three components: regional, synchronous, and temporal transformers. In addition to using a large benchmark database (BETA) toward SSVEP-BCI application to validate model performance, we compared the EEGformer to current state-of-the-art deep learning models using two EEG datasets, which are obtained from our previous study: SJTU emotion EEG dataset (SEED) and a depressive EEG database (DepEEG).ResultsThe experimental results show that the EEGformer achieves the best classification performance across the three EEG datasets, indicating that the rationality of our model architecture and learning EEG characteristics in a unified manner can improve model classification performance.ConclusionEEGformer generalizes well to different EEG datasets, demonstrating our approach can be potentially suitable for providing accurate brain activity classification and being used in different application scenarios, such as SSVEP-based early glaucoma diagnosis, emotion recognition and depression discrimination.

A novel approach for ASD recognition based on graph attention networks

“…Thirdly, the utilization of attention mechanisms enables the identification of influential connections, pinpointing nodes that play crucial roles in the network. Lastly, we adopt a multi-scale perspective by separately extracting features from two graph convolutional layers and concatenating them, providing a comprehensive representation of node features ( Wan et al, 2023b , c ). These combined strategies contribute to the enhanced performance of our algorithm.…”

Understanding the brain with attention: A survey of transformers in brain sciences

“…In addition, some researches mapped EEG signals into feature maps using CNN architecture, implementing selfattention enhancement for latent features. For example, Wan et al [193] proposed a Transformer-based EEG analysis model (EEGformer) for brain activity classification. They deeply convolved EEG signals to obtain channel features and calculated vectors from feature maps along the temporal, regional, and synchronous dimensions.…”

“…In addition, some researches mapped EEG signals into feature maps using CNN architecture, implementing self‐attention enhancement for latent features. For example, Wan et al [193] . proposed a Transformer‐based EEG analysis model (EEGformer) for brain activity classification.…”

Visualization Analysis of Smart Contract Technology Based on Citespace

“…For instance, in emotion recognition, self-supervised learning models using transformers have excelled by capturing relevant features from ECG time-series signals, showcasing superior performance on emotion recognition tasks [53]. Similarly, for human activity recognition, lightweight transformer approaches have been applied successfully using IR-UWB radar [27] and EEG signals [55]. Even WiFi signals have been utilized for indoor human mobility modeling through transformer-based techniques [48].…”