Imagined Speech Brain-Computer Interface: A Task-Oriented Review of Neural Decoding

Area of Science:

• Neuroscience
• Computer Science
• Linguistics

Background:

• Imagined speech decoding is crucial for brain-computer interface (BCI) development, enabling language recovery from neural activity.
• Existing research often treats imagined speech decoding as a unified problem, overlooking significant variations in goals and outputs.

Purpose of the Study:

• To review and categorize recent imagined speech decoding research using a task-oriented framework.
• To analyze how decoding tasks are defined, constrained by output spaces, and expressed through different pathways.

Main Methods:

• Categorization of studies into four task levels: semantic/intent, phoneme/syllable, word, and sentence/language decoding.
• Comparison along dimensions of output-space property (e.g., closed-set) and output pathway.
• Analysis of linguistic granularity and communication objectives.

Main Results:

• Studies exhibit diverse linguistic granularities, ranging from low-bandwidth intent recognition to high-bandwidth text or speech reconstruction.
• Significant differences exist in decoding targets, output constraints, and system output forms across studies.
• A task-oriented framework highlights the heterogeneity of imagined speech decoding.

Conclusions:

• Imagined speech decoding should not be viewed as a homogeneous problem.
• A task-oriented framework offers a clearer basis for comparing diverse studies and advancing communication-oriented BCI research.