Influence of the sinusoidal and Gaussian noises in the estimation of the EEG fractal dimension

EEG signals corresponding to different psychophysiological conditions can be characterized by their fractal dimension (D). The noises present on the recording can affect the estimation of such a dimension. In this work we analyse the behaviour of two D estimators in case of different kinds (gaussian and sinusoidal) and amplitudes of noise. The EEG fractal dimension seems to be strongly compromised by gaussian noise greater than about 3–4%, of the EEG rms value, while a 50Hz noise of about 10% of the EEG rms signal is necessary to produce estimation errors greater than 10%. A dependence on the sampling frequency of the D estimation is also pointed out.