Single trial detection of hand poses in human ECoG using CSP based feature extraction

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Brain-Computer Interfaces (BCI) offer intuitive control but are limited by low-resolution electroencephalogram (EEG) signals.
• Electrocorticographic (ECoG) signals provide higher spatial and temporal resolution, enabling detailed analysis of brain activity.
• High-gamma frequency band (100-500 Hz) in ECoG is crucial for discriminating fine motor tasks.

Purpose of the Study:

• To discriminate between three complex hand movements (open, peace, fist) and an idle state using ECoG signals.
• To evaluate the effectiveness of Common Spatial Patterns (CSP) for feature optimization in BCI.
• To compare CSP-based feature selection with manual feature selection.

Main Methods:

• ECoG signals from two subjects performing hand poses were recorded.
• High-gamma band signals (100-500 Hz) were spatially filtered using CSP.
• Multi-class and two-class Linear Discriminant Analysis (LDA) were employed for classification.
• Manual feature selection was also tested for comparison.

Main Results:

• CSP-based feature selection significantly reduced classification error rates compared to manual selection.
• For discriminating three hand movements, CSP achieved error rates of 7.22% (S1) and 1.17% (S2).
• For discriminating hand movement vs. idle state, CSP achieved error rates of 13.39% (S1) and 2.33% (S2).

Conclusions:

• ECoG signals in the high-gamma band, processed with CSP, are highly effective for decoding complex hand movements.
• CSP is a powerful tool for feature optimization, enhancing BCI performance.
• This approach holds promise for developing independent and intuitively controlled BCIs.