To cite this paper use one of the standards below:
Introduction The evaluation of the sensory perception of food aromas traditionally depends on subjective affective and descriptive methods. Electroencephalography (EEG) offers an objective alternative to measure involuntary neurophysiological responses to olfactory stimuli. Objective To develop and evaluate a Pipeline based on digital processing of EEG signals and artificial intelligence to characterize and discriminate brain responses to natural and synthetic food flavors. Methodology EEG signals were recorded in healthy individuals exposed to aromas. Preprocessing included digital bandpass filtering (Butterworth, 0.4–60 Hz), notch filter (60 Hz), and artifact removal using Discrete Wavelet Transform (Daubechies db4 and Symlet sym5 families). Spectral characteristics (power spectral density via Welch in delta, theta, alpha, beta and gamma bands), statistical and non-linear characteristics were extracted. Machine learning algorithms (PCA, SVM, and 1D Convolutional Neural Networks) were applied combined with Nested LOSO cross validation and permutation tests. Results The combination of bidirectional filtering and normalization by subject (centered mode and deltaC) significantly reduced interindividual variability. The classifiers achieved an accuracy of 61.8% in discriminating between contrasting conditions. In the comparison between natural and synthetic versions of the same aroma (coffee, chocolate, strawberry), the accuracy was between 27% and 40% (p > 0.05 in permutation tests), which mathematically confirms a high overlap of electrophysiological patterns and the neuroaffective equivalence of the stimuli. Conclusions The proposed computational pipeline efficiently processes olfactory EEG signals and confirms the feasibility of neuroscience applied to the aroma industry. The low statistical separability between natural flavors and their synthetic counterparts validates the ability of synthetic formulations to mimic the brain response induced by natural flavors.
Con casi 200.000 artículos publicados, Galoá permite a los académicos compartir y descubrir investigaciones de vanguardia a través de nuestra plataforma de publicación académica optimizada y accesible.
Obtenga más información sobre nuestros productos:
This proceedings is identified by a DOI , for use in citations or bibliographic references. Atención: este no es un DOI para el trabajo y, como tal, no se puede usar en Lattes para identificar un trabajo en particular.
Check the link "How to cite" en la página del papel, para ver cómo citar correctamente el papel