Temporal Feature Extractors in EEG Foundation Models: A Controlled Comparison Including a Pretrained Time-Series Model
For researchers developing EEG foundation models, this work provides a controlled comparison showing that general-purpose time-series models can be reused as frozen temporal extractors, but the benefit depends on the downstream task.
The paper compares three temporal feature extraction strategies (linear, convolutional, and a frozen pretrained time-series foundation model MOMENT) within a unified EEG foundation model. Results show that simple temporal representations perform competitively on motor imagery, while richer temporal modeling benefits emotion recognition, and the pretrained TSFM effectively transfers to EEG.
Electroencephalography (EEG) foundation models aim to learn generalizable representations from large-scale brain recordings. However, the role of temporal feature extractors and whether pretrained time-series foundation models (TSFMs) can be effectively transferred to this setting remains underexplored. We conduct a controlled comparison of three temporal feature extraction strategies, including a linear baseline, a convolutional encoder, and a frozen pretrained TSFM (MOMENT), within a unified EEG foundation model. We evaluate their impact on representation quality using two downstream tasks: motor imagery and emotion recognition. Results reveal different trends across the evaluated benchmarks. On the motor imagery dataset, simple temporal representations perform competitively, whereas the emotion dataset benefits from richer temporal modeling. Although not specifically adapted to EEG, the pretrained TSFM serves as an effective temporal feature extractor, suggesting that general-purpose time-series representations can be transferred as frozen temporal feature extractors within EEG foundation models.