EEG adaptive noise cancellation using information theoretic approach.

Area of Science:

• Biomedical Signal Processing
• Neuroscience Engineering
• Adaptive Filter Theory

Background:

• Electroencephalogram (EEG) signals are susceptible to noise contamination.
• Conventional noise reduction methods often use Mean-Squared Error (MSE), optimizing only the second moment of error distribution.
• Non-Gaussian noise in EEG signals limits the effectiveness of MSE-based approaches.

Purpose of the Study:

• To present an adaptive method for eliminating noise from EEG signals.
• To introduce the error entropy criterion as an alternative to MSE for adaptive filtering.
• To improve the performance of noise elimination in EEG signals.

Main Methods:

• An adaptive filtering method utilizing the error entropy criterion was developed.
• The Minimum Error Entropy (MEE) algorithm was employed to minimize all moments of the error distribution.
• The proposed MEE-based method was compared against conventional MSE-based adaptive algorithms.

Main Results:

• The proposed MEE-based adaptive method demonstrated superior performance in noise elimination compared to MSE-based algorithms.
• Improvements were observed in signal-to-noise ratio (SNR) for the MEE method.
• Reduced steady-state error was achieved using the MEE criterion.

Conclusions:

• Minimizing error entropy offers a more effective approach for noise elimination in EEG signals than MSE.
• The MEE algorithm provides enhanced performance for adaptive filtering of noisy EEG data.
• This method holds promise for improving the quality of EEG recordings.