Cortical Classification with Rhythm Entropy for Error Processing in Cocktail Party Environment Based on Scalp EEG...

Area of Science:

• Neuroscience
• Signal Processing
• Cognitive Science

Background:

• Distinguishing error from correct responses is crucial for understanding cognitive processes.
• Scalp electroencephalography (EEG) offers a non-invasive window into brain activity.
• Auditory attention in complex environments presents significant challenges for signal analysis.

Purpose of the Study:

• To explore a novel feature extraction method based on rhythm entropy for classifying single-trial EEG signals.
• To differentiate error-related EEG signals from correct-response signals during auditory tasks.
• To investigate the neural correlates and information flow patterns associated with error detection.

Main Methods:

• Weighted Minimum Norm Solution Estimation (WMNE) was used to calculate single-trial cortical signals.
• Rhythm entropy was employed as a feature extraction technique.
• Support Vector Machine (SVM) with Leave-One-Out Cross-Validation (LOOCV) was utilized for classification.

Main Results:

• A classification rate of 89.7% was achieved for single-trial EEG signals (≈700 ms).
• Discriminative regions included the medial frontal cortex (MFC), left SMA (lSMA), and right SMA (rSMA).
• Error trials exhibited significantly lower mean entropy values than correct trials in these regions.
• Time-varying network analysis revealed distinct information flow biases: left-bias for error processing and right-bias for correct processing.

Conclusions:

• Rhythm information derived from single cortical signals effectively characterizes error-related EEG signals.
• The findings provide a novel approach for analyzing auditory attention in brain-computer interfaces (BCIs).
• This method enhances our understanding of neural mechanisms underlying error monitoring and cognitive control.