[Evoked Potential Blind Extraction Based on Fractional Lower Order Spatial Time-Frequency Matrix]

Area of Science:

• Signal Processing
• Biomedical Engineering
• Neuroscience

Background:

• Electroencephalograph (EEG) signals often contain impulsive noise with heavy tails and infinite variance, characteristic of stable distributions.
• Traditional signal processing methods struggle with these noise types, impacting the accurate extraction of evoked potentials (EPs).

Purpose of the Study:

• To develop an improved time-frequency distribution and blind source separation method for extracting EPs from EEG signals corrupted by stable noise.
• To introduce a novel fractional lower order spatial time-frequency underdetermined blind source separation (FLO-TF-UBSS) algorithm.

Main Methods:

• Enhanced Wigner-Ville Distribution (WVD) and pseudo WVD (PWVD) using fractional lower order moments (FLO-WVD, FLO-PWVD).
• Proposed fractional lower order spatial time-frequency distribution matrix (FLO-STFM).
• Combined FLO-STFM with time-frequency underdetermined blind source separation (TF-UBSS) to create FLO-TF-UBSS for stable distribution environments.

Main Results:

• Simulations demonstrated that FLO-TF-UBSS effectively extracts EPs from EEG noise, with separated signals closely matching original ones.
• Compared to standard TF-UBSS, FLO-TF-UBSS showed higher correlation coefficients (approaching 1) across varying Generalized Signal-to-Noise Ratios (GSNR) and distribution parameters.
• The proposed method exhibited superior performance over second-order TF-UBSS for EP extraction in noisy EEG.

Conclusions:

• The novel FLO-TF-UBSS algorithm is highly effective for extracting EPs from EEG signals contaminated with stable noise.
• FLO-TF-UBSS offers a significant improvement over existing TF-UBSS methods, particularly in challenging noise conditions.
• This advancement holds promise for more accurate analysis of brain activity in noisy environments.