Error encoding in human speech motor cortex

Area of Science:

• Neuroscience
• Biomedical Engineering
• Speech Science

Background:

• Humans possess self-monitoring abilities for speech production, including error detection.
• Speech neuroprosthesis users can identify errors in decoded output via feedback.
• The presence of neural error detection signals in the speech motor cortex was previously unknown.

Purpose of the Study:

• To investigate the existence of neural error signals within the speech motor cortex.
• To determine if these neural signals can be decoded to enable self-correction in speech neuroprostheses.
• To understand how different types of speech errors are represented neurally.

Main Methods:

• Utilized intracortical recordings from the speech motor cortex during brain-to-text speech neuroprosthesis use.
• Analyzed neural firing rates to identify error-related activity.
• Developed decoding algorithms to detect errors based on neural signals.
• Differentiated neural patterns corresponding to phonemic and semantic error types.

Main Results:

• Demonstrated the presence of neural error signals in the speech motor cortex.
• Achieved up to 86% accuracy in decoding neuroprosthesis errors using these neural signals.
• Identified distinct neural patterns for phonemic versus semantic errors between intended and displayed words.

Conclusions:

• Neural error signals are represented in the speech motor cortex during neuroprosthesis use.
• These signals can be leveraged to improve the accuracy and functionality of speech neuroprostheses.
• Findings suggest new strategies for developing more effective brain-computer interfaces for speech restoration.