Bispectrum Analysis of Noninvasive EEG Signals Discriminates Complex and Natural Grasp Types

Area of Science:

• Neuroscience
• Signal Processing
• Biomedical Engineering

Background:

• Traditional power spectrum analysis in frequency domain analysis omits crucial phase information.
• Understanding neural dynamics during complex motor tasks like grasping requires advanced analytical techniques.

Purpose of the Study:

• To investigate the efficacy of bispectrum analysis in decoding electroencephalography (EEG) data during human grasping movements.
• To evaluate the performance of different classifiers in differentiating grasping motions using bispectrum features.

Main Methods:

• EEG data were collected from five human subjects performing grasping movements.
• Bispectrum analysis was applied to extract magnitude and phase-related features from the EEG data.
• Three distinct classifiers, including Support Vector Machine (SVM), were employed for classification tasks.

Main Results:

• The bispectrum effectively captured phase information crucial for differentiating grasping movements.
• The SVM classifier demonstrated superior performance, achieving 97% accuracy in binary classification (power grasp detection).
• In multiclass tasks, the SVM classifier maintained a high accuracy of 94.93%.

Conclusions:

• The bispectrum is a powerful tool for analyzing neural activity related to complex motor actions.
• The SVM classifier shows significant potential for accurately classifying grasping movements based on EEG bispectrum features.
• This study opens new avenues for understanding neural dynamics and movement decoding.