Purpose: Removal of blink artifacts from the signal using wavelet processing.
Method: Wavelet denoising using the Daubechies 4 (db4) wavelet.
Parameters:
- Decomposition level depends on the sampling frequency:
Fs = 1000 Hz: N = 7 levels
Fs = 500 Hz: N = 6 levels
Otherwise: N = 5 levels - Threshold processing with parameter alpha = 3
- Level-dependent thresholding
How it works:
- Wavelet decomposition of the signal into N levels
- Determination of thresholds for each level based on coefficient statistics
- Thresholding of the coefficients (removing small coefficients corresponding to noise)
- Inverse wavelet transform
- Noise is calculated as the difference:
noise = original_signal - denoised_signal - Result:
signal = original_signal - noise
Use case: Suppression of blink artifacts that distort EEG, especially in frontal channels.
Physiological explanation:
Blinks create short-term, large-amplitude artifacts due to eyelid muscle activity and eye movement. These artifacts have a characteristic shape and temporal structure, which makes it possible to isolate and attenuate them using wavelet transformation while preserving useful neural activity.
Why this matters: Blink artifacts mask weak but important EEG components (for example, the alpha rhythm in the 8–13 Hz range) and can create false changes in the spectrum.