An adaptive joint CCA-ICA method for ocular artifact removal and its application to emotion classification

Area of Science:

• Neuroscience
• Signal Processing
• Machine Learning

Background:

• Electroencephalogram (EEG) signal quality is crucial for understanding neural mechanisms of emotion.
• Ocular artifacts degrade EEG signal-to-noise ratio (SNR), obscuring cognitive components and challenging research.

Purpose of the Study:

• To develop a novel unsupervised learning algorithm for adaptive removal of ocular artifacts from EEG signals.
• To enhance EEG signal quality for improved neuroscience research and brain-computer interface applications.

Main Methods:

• Combined canonical correlation analysis (CCA) and independent component analysis (ICA) for improved source separation.
• Utilized higher-order statistics to pinpoint ocular artifact sources.
• Applied empirical mode decomposition (EMD) and wavelet denoising to correct noisy sources and boost EEG SNR.

Main Results:

• Simulation studies demonstrated significant EEG signal quality improvement across various noise conditions compared to four state-of-the-art methods.
• Real EEG applications confirmed efficient suppression of ocular artifact components.
• Preservation of inherent cognitive processing information enhanced the reliability of Power Spectral Density (PSD) analysis and emotion recognition.

Conclusions:

• The proposed method outperforms existing techniques in EEG recovery and enhances downstream analysis like PSD and emotion recognition.
• This novel approach offers a promising and efficient EEG preprocessing technology for developing brain-computer interfaces, particularly for emotion recognition.