Comparação entre os algoritmos de análise de componentes Independentes e filtragem adaptativa para redução de artefato de piscada de olho em sinais de eletroencefalograma

Abstract

Although the electro-oculogram (EOG) artifact is associated with the inherent human action of blinking, it contaminates and distorts electroencephalogram (EEG) waveforms, impairing the control of Brain-Machine Interfaces (BMI) based on cortical activity. EEGLAB is a widely used toolbox for EOG attenuation in EEG, and has classical techniques for this purpose like the Infomax and SOBI (Second-Order Blind Identification) approaches of the Independent Component Analysis (ICA) method. Despite the wide use of ICA in EOG removal for MCI applications, this technique has difficulty in dealing with bidirectional contamination of the data. In order to mitigate this characteristic of ICA, its variations Wavelet and Adaptive Filtering were evaluated as pre-processing alternatives for MCI. The results showed that the tested methods performed well in reducing EOG by keeping the retrieved EEG as close as possible to the pure EEG signal. However, Adaptive Filtering has shown a slightly superior performance in the analysis of the mean squared error-correlation pair.