Abstract
To reject mechanical artifacts caused by a cochlear implant (CI), we separated electroencephalogram (EEG) in CI users from overlapping signal by using independent component analysis (ICA).We recorded EEG from CI user while they performed an oddball task of auditory stimuli. The participant was asked to compare sequentially presented standard (pure tone of 1 kHz) or target (pure tone of 2 kHz) sounds. The stimulation was presented by a speaker put 30cm from a right sound processor of CI. EEG was obtained at C3, C4 and Cz. ICA were estimated to source signals from observed multivariate -signals by assuming that they are statistically independent from each other and that the subcomponents are non-Gaussian. As the result, the averaged EEG waveform at C4 (close to sound processor) is clearly seen the mechanical artifact from CI. After applying ICA, the artifact component were removed and other components were reconstructed. This result shows that ICA is useful for removal of mechanical artifacts of from CI.
