Characterizing pink and white noise in the human electroencephalogram.

The power spectrum of the human electroencephalogram (EEG) as a function of frequency is a mix of brain oscillations (Osc) (e.g. alpha activity around 10 Hz) and non-Osc or noise of uncertain origin. 'White noise' is uniformly distributed over frequency, while 'pink noise' has an inverse power-frequency relation (power ∝ 1/). Interest in EEG pink noise has been growing, but previous human estimates appear methodologically flawed. We propose a new approach to extract separate valid estimates of pink and white noise from an EEG power spectrum.We use simulated data to demonstrate its effectiveness compared with established procedures, and provide an illustrative example from a new resting eyes-open (EO) and eyes-closed (EC) dataset. The topographic characteristics of the obtained pink and white noise estimates are examined, as is the alpha power in this sample.Valid pink and white noise estimates were successfully obtained for each of our 5400 individual spectra (60 participants × 30 electrodes × 3 conditions/blocks [EO, EC, EO]). The 1/noise had a distinct central scalp topography, and white noise was occipital in distribution, both differing from the parietal topography of the alpha Osc. These differences point to their separate neural origins. EC pink and white noise powers were globally greater than in EO.. This valid estimation of pink and white noise in the human EEG holds promise for more accurate assessment of oscillatory neural activity in both typical and clinical groups, such as those with attention deficits. Further, outside the human EEG, the new methodology can be generalized to remove noise from spectra in many fields of science and technology.