Answers are grounded in this paper and cite their sources.
Purpose: Functional near-infrared spectroscopy (fNIRS) has emerged as a prominent research tool in the fields of brain function analysis, and cognitive load classification due to its ability to objectively reflect cognitive states. However, the variability of fNIRS signals among subjects and the scarcity of data present significant challenges. Methods: To address these issues, this study proposes a cognitive load classification algorithm based on Multimodal Domain Adaptation (MMDA), which comprises an Intra-Modal Domain Adaptation (IMDA) module and a Cross-Modal Domain Adaptation (CMDA) module. The IMDA module aims to mitigate the inter-subject variability of fNIRS signals by extracting shared features among subjects, while the CMDA module leverages electroencephalography (EEG) to align the spatiotemporal information of fNIRS, thereby capturing intrinsic relationships between physiological signals to overcome the limitations in model feature representation caused by the scarcity of fNIRS data. Results: The performance of the proposed model was validated on two publicly available datasets, yielding classification accuracies of 83.51 and 79.56% for cross-subject fNIRS classification tasks, respectively, surpassing existing deep learning and transfer learning models. Conclusion: Notably, EEG signals were utilized solely during the model training phase to enhance the feature representation capabilities of the fNIRS model, thereby facilitating the advancement of fNIRS-based brain-computer interface (BCI) applications.
No takes yet. Share an insight, caveat, or question.
Yu et al. (2025) studied this question.
Synapse has enriched 10 related papers. Consider them for comparative context: