EXTRACTION OF EEG EMOTIVE RESPONSES BY MEANS OF A ICA DATA-DRIVEN APPROACH IN NONSTATIONARY CONDITION

Hypoxia is known to impair cognitive performance, alter perception and produce irreversible brain damage if protracted. To investigate its effects on emotion perception, event related potentials evoked by the presentation of emotional unpleasant aversive and neutral pictures during free breathing and breath holding (BH) were recorded. The experiment was performed by male diving athletes in air. This thesis work includes the complete signal analysis performed on the recorded data, which ranges from the preprocessing and artifacts removal to the development of innovative methods for feature extraction on highly noisy,multivariate and non-stationary processes. By means of a data-driven approach (ICASSO), suitable for disentangling the ERPs components deriving from independent brain sources, brain responses during the free breathing condition were modeled and the changes occurring during BH were described by amplitude-scalings and time-shiftings of the same sources. These components show brain response changes during both the normoxic and the hypoxic phases of BH related to individual brain sources and thus to the functioning of specific brain areas. The 1st component has a posterior localization, no changes during the BH phases were found indicating that the main EEG features of emotional processing are preserved throughout BH; two other components, lateralized, respectively active on the temporo-frontal and on the frontal midline regions, decrease in amplitude during BH with no differences between the normoxic and hypoxic phases. Finally, a component widespread on the scalp but with a frontal prevalence, shows an increasing reduction parallel to the hypoxic trend. The spatial localization of these components is compatible with a set of processing modules that affect automatic and intentional control of attention