Non-linear EEG features for odor pleasantness recognition

Since olfactory sense is gaining ground in multimedia applications, it is important to understand the way odor pleasantness is perceived. Although several studies have explored the way odor pleasantness perception influences the power spectral density of the electroencephalography (EEG) in various brain regions, there are still no studies that investigate the way odor pleasantness perception affects the non-linear temporal variations of EEG. In this study two non-linear metrics are used, namely permutation entropy, and dimension of minimal covers, to explore the possibility of recognizing odor pleasantness perception from the non-linear properties of EEG signals. The results reveal that it is possible to discriminate between pleasant and unpleasant odors from the EEG nonlinear properties, using a Linear Discriminant Analysis classifier with cross-validation.