Preference Detection of the Humanoid Robot Face Based on EEG and Eye Movement
Pengchao Wang,Wei Mu,Gege Zhan,Aiping Wang,Zuoting Song,Tao Fang,Xueze Zhang,Junkongshuai Wang,Lan Niu,Jianxiong Bin,Lihua Zhang,Jie Jia,Xiaoyang Kang
DOI: https://doi.org/10.1007/s00521-024-09765-0
2024-01-01
Neural Computing and Applications
Abstract:The face of a humanoid robot can affect the user experience, and the detection of face preference is particularly important. Preference detection belongs to a branch of emotion recognition that has received much attention from researchers. Most of the previous preference detection studies have been conducted based on a single modality. In this paper, we detect face preferences of humanoid robots based on electroencephalogram (EEG) signals and eye movement signals for single modality, canonical correlation analysis fusion modality, and bimodal deep autoencoder (BDAE) fusion modality, respectively. We validated the theory of frontal asymmetry by analyzing the preference patterns of EEG and found that participants had higher alpha wave energy for preference faces. In addition, hidden preferences extracted by EEG signals were better classified than preferences from participants' subjective feedback, and also, the classification performance of eye movement data was improved. Finally, experimental results showed that BDAE multimodal fusion using frontal alpha and beta power spectral densities and eye movement information as features performed best, with the highest average accuracy of 83.13