A portable affective computing system for identifying mate preference

Abstract

Recognizing individuals' preference states for potential romantic partners based on electroencephalography (EEG) signals holds significant application value in enhancing matchmaking success rates and preventing romantic fraud. Despite some progress has been made in this field, challenges such as high-dimensional feature space and channel redundancy have hindered the translation of this technology. This paper explores how to find out the most discriminative EEG features and channels to reduce data collection, thereby enhancing the convenience and translational potential of EEG-based romantic attraction recognition systems. For this purpose, the present study initially devised an interesting simulated mate selection experiment to gather the required data. Subsequently, EEG features were extracted from different dimensions, encompassing frequency band power and asymmetry index features. Additionally, recursive feature elimination (RFE) algorithm and frequency-based feature subset integration algorithm (FFSI) were integrated as a novel approach for EEG feature and channel selection. Lastly, random forest classifier (RFC) was employed to identify individuals' preference states for potential romantic partners to validate the performance of RFE-FFSI. Experimental results demonstrate that the features in the optimal feature subset primarily come from electrodes located on the frontal lobe, and an average classification accuracy of 88.09% can be achieved based on this optimal feature subset.