Abstract:EEG signals play a crucial role in assessing cognitive load, which is a key element in ensuring the secure operation of human–computer interaction systems. However, the variability of EEG signals across different subjects poses a challenge in applying the pre-trained cognitive load assessment model to new subjects. Moreover, previous domain adaptation research has primarily focused on developing complex network architectures to learn more domain-invariant features, overlooking the noise introduced by pseudo-labels and the challenges posed by domain migration problems. Therefore, this study proposes a novel cross-attention swin-transformer network for cross-subject cognitive load assessment, which achieves inter-domain feature alignment through parameter sharing in cross attention mechanism without using pseudo-labels, and utilizes maximum mean discrepancy (MMD) to measure the difference between the feature distributions of the source and target domains, further promoting feature alignment between domains. This method aims to leverage the advantages of cross-attention mechanism and MMD to better mitigate individual differences among subjects in cross-subject cognitive workload assessment. To validate the classification performance of the proposed network, two datasets of image recognition task and N-back task were employed for testing. Results show that, the proposed model outperformed advanced methods with cross-subject classification results of 88.13% and 81.27% on the on local and public datasets. The ablation experiment results reveal that using either the cross-attention mechanism or the MMD strategy alone improves cross-subject classification performance by 2.11% and 2.95% on the local dataset, respectively. Furthermore, the results of the EEG features distribution differences between all subjects before and after network training showed a significant reduction in feature distribution differences between subjects, further confirming the network's effectiveness in minimizing inter-subject differences.

Cross-subject EEG Channel Optimization by Domain Adversarial Sparse Learning Model

Cross-Subject EEG Signal Recognition Using Deep Domain Adaptation Network

An EEG-Based Transfer Learning Method for Cross-Subject Fatigue Mental State Prediction

Avoiding subject-specific model selection via highway networks in EEG signals

Reducing the Subject Variability of EEG Signals with Adversarial Domain Generalization

Lightweight Source-Free Domain Adaptation based on Adaptive Euclidean Alignment for Brain-Computer Interfaces

Cross-Subject Cognitive Workload Recognition Based on EEG and Deep Domain Adaptation

Sparse Logistic Regression-Based EEG Channel Optimization Algorithm for Improved Universality across Participants

An improved multi-source domain adaptation network for inter-subject mental fatigue detection based on DANN

Label-Based Alignment Multi-Source Domain Adaptation for Cross-Subject EEG Fatigue Mental State Evaluation

Dynamic threshold distribution domain adaptation network: a cross-subject fatigue recognition method based on EEG signals

Less is More: Brain Functional Connectivity Empowered Generalisable Intention Classification with Task-relevant Channel Selection

Common Spatial Generative Adversarial Networks based EEG Data Augmentation for Cross-Subject Brain-Computer Interface

A Deep Subdomain Associate Adaptation Network for Cross-Session and Cross-Subject Eeg Emotion Recognition

BDAN-SPD: A Brain Decoding Adversarial Network Guided by Spatiotemporal Pattern Differences for Cross-Subject MI-BCI

BDAN: Mitigating Temporal Difference Across Electrodes in Cross-Subject Motor Imagery Classification via Generative Bridging Domain

Deep Representation-Based Domain Adaptation for Nonstationary EEG Classification

Prototypical Contrastive Domain Adaptation Network for Nonstationary EEG Classification

Towards Cross-Brain Computer Interface: A Prototype-Supervised Adversarial Transfer Learning Approach with Multiple Sources

A Cross-Attention-Based Class Alignment Network for Cross-Subject EEG Classification in a Heterogeneous Space

A cross-attention swin transformer network for EEG-based subject-independent cognitive load assessment