DIVA Meets EEG: Model Validation Using Formant-Shift Reflex.

Area of Science:

• Neuroscience
• Speech Science
• Computational Modeling

Background:

• The Directions into Velocities of Articulators (DIVA) model explains speech production and acquisition.
• Previous validation used functional magnetic resonance imaging (fMRI), which has limitations in temporal resolution.

Purpose of the Study:

• Introduce DIVA-EEG, an extension of the DIVA model using electroencephalography (EEG).
• Leverage EEG's high temporal resolution and accessibility for neurocomputational speech modeling.
• Provide physiological validation for the DIVA model using EEG.

Main Methods:

• Derived EEG-like signals from DIVA model equations.
• Generated synthetic EEG data simulating syllable utterances with auditory feedback perturbations.
• Acquired empirical EEG data from 30 participants with typical voices during altered auditory feedback.
• Compared synthetic EEG-derived cortical maps with empirical brain activity.

Main Results:

• Synthetic EEG cortical activation maps closely matched the original DIVA model's maps.
• Empirical brain activity maps significantly overlapped with DIVA-EEG predictions.
• Demonstrated the feasibility of using EEG for neurocomputational speech modeling.

Conclusions:

• DIVA-EEG offers a viable, high-temporal-resolution alternative to fMRI for validating neurocomputational speech models.
• This framework supports the development of comprehensive neurocomputational tools for speech and vocal disorders.
• Lays the groundwork for model-driven, personalized interventions for speech impairments.