• auditory scene analysis
• inverse problem
• ERS
• ERD
• multitapering
• cocktail party effect

In this study brain's behavior during a typical cocktail party situation has been investigated. To this purpose, stereo and mono complex auditory scenes composed of a mixture of human voice and natural environmental sounds were used for the stimulation; as experimental task, subjects were asked to focus their attention on the voice or the background noise. Thus, the experimental conditions were: mono attention to voice (mav), mono attention to environment (mae), stereo attention to voice (sav), stereo attention to environment (sae). Magnetoencephalography (MEG) has been used as neuroimaging modality for its high temporal resolution and its property to reflect dynamics of neural activity directly.
MEG data they were checked for physiological and electronics artifacts by combining artifacts correction and artifacts rejection strategies. They were first cleaned from artifacts deriving from eyes blinks, eyes movements and heart activity by using ICA; then, a visual inspection of the data trial by trial was performed in order to remove trials contaminated with artifacts caused by electronics or head movements.
Three main analysis of artifacts-free data were performed: an analysis of the time course and topographies of the evoked fields at the sensor level, an analysis of the sources underlying the observed evoked fields, and a time-frequency analysis of the MEG signal at the sensor level. For each condition of each subject in the study, evoked fields were obtained by averaging all trials left after the artifact rejection procedure; they were baseline corrected with respect to a prestimulus period, transformed to planar gradients, and pooled across subjects to obtain grand-averaged planar evoked fields. For a raw assessment of auditory activity, time courses of sensors located over the auditory areas were averaged separately for each brain hemisphere. Consistent with the typical magnetic auditory evoked response, signals in the auditory channels presented two transient responses, peaked respectively around 50 ms (M50) and 100 ms (M100) after stimulus onset, followed by a response sustained over several hundreds of milliseconds. To localize the current sources underlying the evoked fields without a priori assumptions about their number and location, a cortically constrained depth-weighted noise-normalized minimum norm estimate (MNE) was performed.In order to obtain statistical maps of the main and differential effects, single trial data from each subject and each experimental condition were projected to the cortex source space. Main and dierential effects in a given time latency were evaluated via t-tests at each vertex of the source space. MNE localized sources of activity in left and right auditory cortex, though for all conditions the activation was stronger in the right hemisphere. The multi-subject statistical fixed-effects analysis revealed no signicant differential effects between sav and sae conditions. By contrast, statistical comparison of mav and mae conditions in the interval 150 ms - 360 ms was able to identify one significant region (p< 0.009, uncorrected) for the distribution of the differential effects. This source was located in the right auditory cortex and exhibited a negative difference between mae and mav conditions (mav>mae).
The sensor-level analysis of the induced response was performed separately for two frequency ranges: 4-30 Hz and 30 - 100 Hz. Time-frequency representations (TFRs) of signal power were computed for each trial and averaged. The obtained TFRs were then averaged across subjects. In order to assess the phase relation of the observed change in power to the stimulus onset, the inter-trial coherence was computed.