ICA-based positive semidefinite matrix templates for eye-blink artifact removal from EEG signal with single-electrode
Suguru Kanoga, Yasue Mitsukura · 2015
Complete artifact removal of eye-blink for electroencephalographic (EEG) signal is generally-regarded as important process in EEG signal analysis. No numerical approach of eye-blink artifact removal for single-channel EEG signal in time domain is developed. In this paper, we propose a time domain eye-blink artifact removal method for single-channel EEG signal using independent component analysis (ICA)-based templates. ICA and positive semidefinite tensor factorization (PSDTF) are employed in our proposed method. As a result, we obtain high signal-to-noise ratio (15.03dB) and low mean square error (25.76). Moreover, the optimal number of iterations on multiplicative update rules in PSDTF is about 20 to 30. A useful time domain eye-blink artifact removal method for single-channel EEG signal was proposed by using ICA-based PSD matrix templates. Improvement of eye-blink artifact removal accuracy and reduction of computational cost are our main future works.