Decoding covert speech for intuitive control of brain-computer interfaces based on single-trial EEG: a feasibility...

Area of Science:

• Neuroscience
• Biomedical Engineering
• Rehabilitation Technology

Background:

• Brain-computer interfaces (BCIs) offer alternative control for individuals with severe motor impairments.
• Motor imagery (MI) is a common BCI control signal, but not universally effective.
• Existing BCIs often lack sufficient control options for complex tasks.

Purpose of the Study:

• To investigate the feasibility of decoding covert speech from single-trial EEG.
• To compare and combine covert speech decoding with motor imagery (MI) for BCI control.
• To explore a multimodal BCI system for enhanced usability.

Main Methods:

• EEG data recorded from seven healthy subjects during covert/overt speech and motor tasks (imagery/execution).
• Temporal and spectral features extracted from EEG epochs.
• Random forest classifier used for decoding speech and movement intentions.

Main Results:

• Average classification accuracy for covert speech was 67% ± 9%, and for overt speech was 75% ± 7%.
• Speech decoding accuracy was 5-10% lower than movement classification.
• Combined movement-speech decoder achieved 61% ± 9% (covert) and 67% ± 7% (overt) accuracy, enabling more control classes.

Conclusions:

• Covert speech decoding from EEG is feasible for BCI control.
• This method provides a potential alternative or supplement to motor imagery.
• Decoding covert speech is a step towards developing more versatile multimodal BCI systems.