Detecting Silent Vocalizations in a Locked-In Subject

Area of Science:

• Neuroscience
• Biomedical Engineering
• Speech Science

Background:

• Decoding neural signals is key for brain-computer interfaces (BCIs).
• Silent vocalization onset detection is needed to improve speech synthesis in BCIs for locked-in patients.
• Current BCIs require enhanced methods for real-time neural signal interpretation.

Purpose of the Study:

• To identify neural correlates of vocalization onset during silent speech.
• To investigate the utility of lower beta band peaks for detecting speech initiation.
• To assess the potential of these neural markers for improving BCIs.

Main Methods:

• Cortical recordings were performed on a mute, paralyzed subject during attempted inner speech.
• Analysis focused on lower beta band power spectrum density (PSD) peaks exceeding a 0.2% increase.
• Electroencephalography (EEG) was used in speaking subjects to corroborate findings over speech areas.

Main Results:

• Significant lower beta band peaks were consistently observed at the estimated onset of silent speech.
• Single unit firing correlations increased post-peak, suggesting a link to motor execution.
• Control studies indicated these peaks relate to articulator movement, not higher cognitive speech processes.

Conclusions:

• A sharp peak in lower beta band activity signifies the onset of both silent and overt speech.
• This discovery is a significant advancement for developing speech prostheses.
• These lower beta band peaks offer potential for online applications to aid speech decoding in BCIs for mute individuals.