Transfer Learning Via Distributed Brain Recordings Enables Reliable Speech Decoding Aditya Singh Tessy Thomas Jinlong Li Greg Hickok Xaq Pitkow Nitin Tandon by Aditya Singh & Tessy Thomas & Jinlong Li & Greg Hickok & Xaq Pitkow & Nitin Tandon instant download after payment.

Speech brain-computer interfaces (BCIs) combine neural recordings with largelanguage models to achieve real-time intelligible speech. However, thesedecoders rely on dense, intact cortical coverage and are challenging to scale1234567890():,;1234567890():,;across individuals with heterogeneous brain organization. To derive scalabletransfer learning strategies for neural speech decoding, we used minimallyinvasive stereo-electroencephalography recordings in a large cohort performing a demanding speech motor task. A sequence-to-sequence model enableddecoding of variable-length phonemic sequences prior to and during articulation. This enabled development of a cross-subject transfer learning frameworkto isolate shared latent manifolds while enabling individual model initialization.The group-derived decoder significantly outperformed models trained onindividual data alone, enabling decoding robustness despite variable coverageand activation. These results highlight a pathway toward generalizable neuralprostheses for speech and language disorders by leveraging large-scale intracranial datasets with distributed spatial sampling and shared task demands.