CIRE: A Chinese EEG Dataset for decoding speech intention modulated by prosodic emotion.

Area of Science:

• Cognitive Neuroscience
• Neurotechnology
• Speech Processing

Background:

• Decoding speech intention is crucial for brain-computer interface (BCI) advancement.
• Existing datasets lack diversity in decoding speech intentions from identical text.
• Prosodic emotion significantly influences the interpretation of spoken language.

Purpose of the Study:

• To introduce the CIRE dataset for spoken language interaction intention.
• To facilitate research on decoding nuanced speech intentions.
• To support advancements in BCI and cognitive neuroscience.

Main Methods:

• Collected high-density (128-channel) EEG data from 38 participants.
• Utilized Wav2vec2-derived acoustic embeddings for speech stimuli.
• Applied signal processing, cognitive analysis, and machine learning for validation.

Main Results:

• Achieved a 68.2% cross-subject classification accuracy with a baseline model.
• Identified interpretable neurophysiological correlates for intention differences.
• Demonstrated the dataset's utility in cognitive neuroscience and BCI.

Conclusions:

• The CIRE dataset addresses the need for diverse speech intention data.
• High-density EEG data supports cognitive neuroscience and speech BCI applications.
• Findings contribute to brain-inspired algorithms and BCI development.