Extracting ERP by combination of subspace method and lift wavelet transform

Area of Science:

• Neuroscience
• Signal Processing
• Biomedical Engineering

Background:

• Scalp-recorded Event Related Potentials (ERPs) are often contaminated by background Electroencephalography (EEG) noise.
• Standard wavelet transforms are ineffective at removing this colored EEG noise due to its complex spectral characteristics.

Purpose of the Study:

• To develop an improved method for reducing EEG noise in ERP recordings.
• To enhance the extraction of Event Related Potentials (ERPs) by minimizing background noise.
• To reduce the number of trials required for reliable ERP analysis.

Main Methods:

• A hybrid approach combining subspace methods (Singular Value Decomposition - SVD) and lift wavelet transform was proposed.
• Singular Value Decomposition (SVD) was used to estimate the signal subspace and pre-whiten the colored EEG noise.
• The pre-denoised signal was then processed using lift wavelet transform for enhanced ERP extraction.

Main Results:

• The combined subspace method and lift wavelet transform demonstrated superior performance in noise reduction compared to individual methods.
• The proposed approach effectively reduced background EEG noise, leading to cleaner ERP signals.
• Simulation results confirmed the enhanced capability of the combined technique.

Conclusions:

• The combination of subspace methods and lift wavelet transform offers a powerful solution for denoising ERPs.
• This novel approach significantly improves the efficiency and accuracy of extracting brain signals from noisy EEG data.
• The method holds promise for advancing neurophysiological research by enabling more robust ERP analysis with fewer trials.