Search papers, labs, and topics across Lattice.
This paper explores transfer learning for MEG-based speech decoding, addressing the challenge of data scarcity in brain-computer interfaces. A Conformer model is pre-trained on 50 hours of single-subject MEG listening data and then fine-tuned on 5 minutes of data per subject for both speech perception and production tasks. The results demonstrate that transfer learning improves both in-task (1-4%) and cross-task (5-6%) accuracy, and that models trained on speech production can decode passive listening, indicating shared neural representations.
Unlock data-efficient speech BCIs: Transfer learning from MEG speech perception to production boosts decoding accuracy even with just 5 minutes of subject-specific data.
Data-efficient neural decoding is a central challenge for speech brain-computer interfaces. We present the first demonstration of transfer learning and cross-task decoding for MEG-based speech models spanning perception and production. We pre-train a Conformer-based model on 50 hours of single-subject listening data and fine-tune on just 5 minutes per subject across 18 participants. Transfer learning yields consistent improvements, with in-task accuracy gains of 1-4% and larger cross-task gains of up to 5-6%. Not only does pre-training improve performance within each task, but it also enables reliable cross-task decoding between perception and production. Critically, models trained on speech production decode passive listening above chance, confirming that learned representations reflect shared neural processes rather than task-specific motor activity.
