Chisco: An EEG-based BCI dataset for decoding of imagined speech

Area of Science:

• Neuroscience and Biomedical Engineering
• Focus on Brain-Computer Interfaces (BCIs) and neural decoding

Background:

• Deep learning advancements enhance Brain-Computer Interfaces (BCIs) and neural decoding accuracy.
• Decoding imagined speech is a key area of interest due to its potential for 'mind reading' applications.
• Previous research is limited by a lack of electroencephalography (EEG) datasets for imagined speech.

Purpose of the Study:

• To introduce the Chinese Imagined Speech Corpus (Chisco), a novel and extensive dataset for imagined speech research.
• To provide a valuable resource for advancing neural language decoding and BCI development.
• To overcome the limitations posed by the scarcity of imagined speech EEG data.

Main Methods:

• Collected high-density EEG recordings from healthy adults imagining speech.
• Dataset includes over 20,000 sentences and 6,000 everyday phrases across 39 semantic categories.
• Each subject's data exceeds 900 minutes, establishing a new benchmark for individual neural language datasets.

Main Results:

• The Chisco dataset is the largest per-individual resource currently available for neural language decoding.
• The comprehensive nature of the stimuli (phrases across semantic categories) allows for broad application.
• The dataset facilitates in-depth analysis of neural patterns associated with imagined speech.

Conclusions:

• Chisco represents a significant contribution to the field of BCIs and neural decoding.
• The availability of this large-scale imagined speech EEG dataset will accelerate research and development.
• This resource is expected to foster the creation of more intuitive and effective BCIs.