ASPEN: Spectral-Temporal Fusion for Cross-Subject Brain Decoding
arXiv:2602.16147v1 Announce Type: new Abstract: Cross-subject generalization in EEG-based brain-computer interfaces (BCIs) remains challenging due to individual variability in neural signals. We investigate whether spectral representations offer more stable features for cross-subject transfer than temporal waveforms. Through correlation analyses across three EEG paradigms (SSVEP, P300, and Motor Imagery), we find that spectral features exhibit consistently higher cross-subject similarity than temporal signals. Motivated by this observation, we introduce ASPEN, a hybrid architecture that combines spectral and temporal feature streams via multiplicative fusion, requiring cross-modal agreement for features to propagate. Experiments across six benchmark datasets reveal that ASPEN is able to dynamically achieve the optimal spectral-temporal balance depending on the paradigm. ASPEN achieves the best unseen-subject accuracy on three of six datasets and competitive performance on others, demo
arXiv:2602.16147v1 Announce Type: new Abstract: Cross-subject generalization in EEG-based brain-computer interfaces (BCIs) remains challenging due to individual variability in neural signals. We investigate whether spectral representations offer more stable features for cross-subject transfer than temporal waveforms. Through correlation analyses across three EEG paradigms (SSVEP, P300, and Motor Imagery), we find that spectral features exhibit consistently higher cross-subject similarity than temporal signals. Motivated by this observation, we introduce ASPEN, a hybrid architecture that combines spectral and temporal feature streams via multiplicative fusion, requiring cross-modal agreement for features to propagate. Experiments across six benchmark datasets reveal that ASPEN is able to dynamically achieve the optimal spectral-temporal balance depending on the paradigm. ASPEN achieves the best unseen-subject accuracy on three of six datasets and competitive performance on others, demonstrating that multiplicative multimodal fusion enables effective cross-subject generalization.
Executive Summary
This research introduces ASPEN, a novel hybrid architecture for cross-subject brain-computer interfaces (BCIs) that leverages spectral and temporal feature streams to achieve optimal performance. Through correlation analyses and experiments on six benchmark datasets, the study demonstrates that ASPEN's multiplicative multimodal fusion enables effective cross-subject generalization. The results show that ASPEN achieves the best or competitive performance on five out of six datasets, indicating its potential for real-world applications. However, the study's reliance on EEG data and limited exploration of other modalities may limit its generalizability to other BCI applications.
Key Points
- ▸ ASPEN is a hybrid architecture that combines spectral and temporal feature streams for cross-subject brain-computer interfaces (BCIs).
- ▸ The study demonstrates that spectral features exhibit consistently higher cross-subject similarity than temporal signals in EEG-based BCIs.
- ▸ Experiments on six benchmark datasets show that ASPEN achieves the best or competitive performance on five out of six datasets.
Merits
Strength
ASPEN's multiplicative multimodal fusion enables effective cross-subject generalization, achieving optimal performance on five out of six datasets.
Methodological innovation
The study introduces a novel hybrid architecture that combines spectral and temporal feature streams, offering a promising approach for BCI applications.
Empirical validity
The results of the study are supported by experiments on six benchmark datasets, providing strong empirical evidence for the effectiveness of ASPEN.
Demerits
Limitation
The study relies on EEG data, which may not be generalizable to other BCI applications or modalities.
Lack of exploration
The study does not explore the potential use of other modalities or the generalizability of ASPEN to other BCI applications.
Expert Commentary
The study introduces a novel hybrid architecture for cross-subject BCIs, leveraging spectral and temporal feature streams to achieve optimal performance. The results demonstrate the effectiveness of ASPEN's multiplicative multimodal fusion, achieving the best or competitive performance on five out of six datasets. However, the study's reliance on EEG data and limited exploration of other modalities may limit its generalizability to other BCI applications. Future research should explore the potential use of other modalities or the generalizability of ASPEN to other BCI applications. Additionally, the study's findings may inform the development of policies or guidelines for the use of BCIs in various applications.
Recommendations
- ✓ Future research should explore the potential use of other modalities or the generalizability of ASPEN to other BCI applications.
- ✓ Further studies should investigate the practical implications of ASPEN's performance and explore its potential applications in various fields, such as healthcare, education, or rehabilitation.
