Removal of Eye Blink Artifacts From EEG Signals Using Sparsity

Area of Science:

• Neuroscience
• Biomedical Engineering
• Signal Processing

Background:

• Electroencephalography (EEG) signals are frequently contaminated by eye blink (EB) artifacts, leading to erroneous activation in brain-computer interface (BCI) systems.
• Existing artifact removal methods, such as ICA and wavelet transformation, are computationally intensive and can cause significant information loss, hindering real-time BCI applications.

Purpose of the Study:

• To investigate and evaluate sparsity-based methods for removing eye blink artifacts from EEG signals.
• To address the limitations of existing methods, focusing on computational efficiency and minimal information loss for online BCI development.

Main Methods:

• Evaluated two sparsity-based techniques: Morphological Component Analysis (MCA) using predefined dictionaries (Dirac, DCT) and K-SVD using a learned dictionary from EEG data.
• Experiments were conducted using both synthetic and real EEG data to assess algorithm performance.

Main Results:

• The K-SVD algorithm demonstrated superior performance in suppressing eye blink artifacts compared to the MCA technique.
• Sparsity-based algorithms achieved performance comparable to the state-of-the-art FORCe method.
• Accurate elimination of EB artifacts was achieved using over-complete dictionaries without computationally expensive algorithms.

Conclusions:

• Sparsity-based methods, especially K-SVD with learned dictionaries, offer an effective and computationally efficient solution for eye blink artifact removal in EEG.
• These methods are suitable for developing online BCI systems due to their accuracy and low computational cost.