Search papers, labs, and topics across Lattice.
This study investigates the relationship between neural representations of vocalized, mimed, and imagined speech using stereotactic EEG recordings. Linear spectrogram reconstruction models were trained for each condition and evaluated for cross-condition generalization, revealing shared speech representations across vocalized, mimed, and imagined speech. Stimulus-level discriminability analysis further confirmed the preservation of stimulus-specific structure across conditions, with linear models outperforming nonlinear neural networks in this regard.
Linear models of neural activity surprisingly decode imagined speech better than nonlinear neural networks, suggesting simpler representations than expected.
We investigated the relationship among neural representations of vocalized, mimed, and imagined speech recorded using publicly available stereotactic EEG recordings. Most prior studies have focused on decoding speech responses within each condition separately. Here, instead, we explore how responses across conditions relate by training linear spectrogram reconstruction models for each condition and evaluate their generalization across conditions. We demonstrate that linear decoders trained on one condition generally transfer successfully to others, implying shared speech representations. This commonality was assessed with stimulus-level discriminability by performing a rank-based analysis demonstrating preservation of stimulus-specific structure in both within- and across-conditions. Finally, we compared linear reconstructions to those from a nonlinear neural network. While both exhibited cross-condition transfer, linear models achieve superior stimulus-level discriminability.
