Speech imagery brain-computer interfaces: a systematic literature review.

Area of Science:

• Neuroscience
• Cognitive Science
• Brain-Computer Interfaces

Background:

• Speech Imagery (SI) involves the mental experience of hearing speech, potentially underpinning verbal thought.
• SI may involve kinesthetic sensations similar to overt speech production without articulation.
• SI is explored as an intuitive paradigm for Brain-Computer Interfaces (BCI).

Purpose of the Study:

• To review the methodologies and outcomes of decoding Speech Imagery (SI) from neural signals for BCI applications.
• To identify trends, challenges, and the current state-of-the-art in SI-BCI decoding pipelines.
• To discuss future research directions in SI-BCI development.

Main Methods:

• Systematic review following PRISMA guidelines.
• Searched Google Scholar and PubMed for peer-reviewed reports on SI decoding from neural activity.
• Selected 104 reports for analysis, focusing on neuroimaging modalities, signal processing, feature extraction, and decoding efficacy (Information Transfer Rates).

Main Results:

• Significant growth in SI decoding research over the past 20 years.
• High preference for Deep Learning models in signal processing and feature extraction.
• Fewer than 6% of studies reported real-time decoding, with most focusing on offline analysis.
• Varied approaches and outcomes complicate the identification of the current state-of-the-art.

Conclusions:

• Speech Imagery (SI) presents an attractive and increasingly researched BCI paradigm.
• Methodological trends show a preference for Deep Learning, but real-time decoding remains a significant hurdle.
• Further research is needed to establish robust, real-time SI-BCI systems and define the field's state-of-the-art.