Single-channel EEG signal extraction based on DWT, CEEMDAN, and ICA method

Area of Science:

• Neuroscience
• Signal Processing
• Biomedical Engineering

Background:

• Portable electroencephalogram (EEG) monitoring for applications like anesthesia depth, emotional state, and sleep monitoring requires user-friendly equipment.
• Removing electrooculogram (EOG) artifacts is challenging with limited EEG channels due to the overcomplete problem, where observed signals are fewer than source signals.
• Standard Independent Component Analysis (ICA) fails with overcomplete data as it requires fewer basis vectors than input dimensions.

Purpose of the Study:

• To propose a novel method for removing EOG artifacts from EEG signals, specifically addressing the challenges posed by overcomplete data and modal aliasing.
• To develop a robust artifact removal technique suitable for portable EEG monitoring systems.
• To improve the accuracy and reliability of EEG signal analysis in practical, real-world scenarios.

Main Methods:

• A hybrid approach combining Discrete Wavelet Transform (DWT), Complete Ensemble Empirical Mode Decomposition with Adaptive Noise (CEEMDAN), and Independent Component Analysis (ICA).
• DWT is applied first, followed by CEEMDAN to decompose the signal, resolving overcomplete and modal aliasing issues.
• ICA is then used to identify and remove EOG artifact components based on their sample entropy values.

Main Results:

• The proposed method effectively removes EOG artifacts from EEG signals.
• The technique successfully addresses the overcomplete problem inherent in low-channel count EEG acquisition.
• Modal aliasing issues, often caused by signal anomalies and noise, are also resolved by the CEEMDAN component of the method.

Conclusions:

• The novel EEG artifact removal method integrating DWT, CEEMDAN, and ICA offers a robust solution for portable monitoring applications.
• This approach overcomes limitations of traditional methods like standard ICA in handling overcomplete and noisy EEG data.
• The findings suggest improved feasibility and accuracy for real-time EEG analysis in diverse clinical and personal monitoring settings.