Search papers, labs, and topics across Lattice.
LibriBrain100 is a comprehensive MEG dataset designed for speech decoding, featuring over 100 hours of high-quality recordings from subjects listening to naturalistic speech, significantly surpassing previous datasets in both size and depth. The dataset includes approximately 80 hours from a single subject, setting a new benchmark for within-subject neural data, and demonstrates state-of-the-art performance on a word-classification task, showcasing the effectiveness of deep, individualized data. Additionally, the dataset supports broader multi-subject data collection, revealing that supervised finetuning can effectively enhance model performance even with limited per-subject data, thus paving the way for advancements in non-invasive brain-computer interfaces.
Achieving state-of-the-art performance in speech decoding with a dataset that features 80 hours of deep, within-subject MEG data sets a new standard for neural data quality and quantity.
We introduce LibriBrain100, a large-scale MEG dataset for speech decoding designed from the ground up for reproducible, standardised evaluation. LibriBrain100 more than doubles the size of the original LibriBrain release, resulting in over 100 hours of high-quality MEG acquired while subjects listened to naturalistic continuous speech. With $\sim$80 hours from a single subject, LibriBrain100 sets a new record for deep, within-subject neural data (8$\times$ more than the next comparable dataset and roughly 80$\times$ more than other datasets). To demonstrate the payoff of this depth-first design, we evaluate on a word-classification benchmark---an increasingly well-established stepping stone towards the open challenge of noninvasive brain-to-text decoding. Using an existing decoding model, we achieve state-of-the-art performance---validating both the quality of the recordings and the value of within-subject data at scale. Because collecting 80 hours of data per user is impractical for real-world applications, we also collected $\sim$40 minutes of additional data from each of 32 subjects. Using the same word-classification benchmark, we demonstrate the value of broad multi-subject data: supervised finetuning of a pre-trained model can substantially compensate for limited per-subject data. We provide standard train, validation, and test splits, all reproducible through an open-sourced Python library that supports easy downloading, optional preprocessing, and data loading for common deep learning frameworks. In addition, the dataset and evaluation infrastructure are being released alongside an open machine-learning competition with a public leaderboard for standardised benchmarking. Ultimately, our hope is that LibriBrain100 will accelerate progress towards practical non-invasive brain-computer interfaces, capable of restoring communication to people living with severe paralysis.
Stanford HAI