EEG Feature Selection for Subjective Preference

Abstract

© 2020 The Society of Instrument and Control Engineers - SICE. The quantitative-based human behavior finding is one of the most challenging problem to access the human preference in neuromarketing. The humans' subconscious response to a decision making relevant to the process of brain electrical signal using electroencephalogram (EEG). The EEG signals are related to the positions of the electrode in the international 10 - 20 system. In this research, the EEG of five participants are recorded while they are choosing the product which is preferred and unpreferred from the nearly identical composition and flavor of the two brands of snack. The signal was normalized and extracted using the power spectral density (PSD) and asymmetry index to observe behavioral differences between the two cerebral hemispheres. According to the experiments, the left frontal with asymmetry index can potentially illustrate the preferred product. To further emphasize, by using the root mean square error calculated from the support vector machine (SVM), the results show that the error in the group using asymmetry feature is less than using the traditional normative PSD feature. According to these results, a systematic human preference prediction can be potentially done using the EEG.