Autoencoder-enabled eye-tracking data analytics for objective assessment of user preference in humanoid robot appearance design
Fan Li,Chun-Hsien Chen,Yisi Liu,Danni Chang,Jian Cui,Olga Sourina
DOI: https://doi.org/10.1016/j.eswa.2024.123857
IF: 8.5
2024-03-31
Expert Systems with Applications
Abstract:Humanoid robots have been extensively utilized in service industries to provide information and product delivery through direct interactions with users. As the design of humanoid robot appearance significantly impacts human-robot interactions, it is crucial to assess user preference towards it. Traditional evaluation tools, such as surveys, field observations, and interviews, are often time-consuming and subjective. Therefore, this study aims to develop a novel eye-tracking-based assessment tool to investigate user preference towards humanoid robot appearance design. We analyze the critical factors influencing user preference from two perspectives: the attributes of robot appearance and users' selective attention distribution. Accordingly, we propose an integrated machine learning method, combining an autoencoder neural network with a support vector machine to handle the collected visual data. This method, named ASVM, extracts several novel indicators from the eye-tracking data via an unsupervised autoencoder neural network and manual entropy analysis. The proposed ASVM achieves an accuracy of 91%, outperforming other classical machine learning methods, including decision tree, naive Bayes, and support vector machine. ASVM can objectively assess user preference towards humanoid robot appearance design with high time resolution. Furthermore, it can enhance humanoid robot design by revealing the visual attention distribution in assessing robot appearance.
computer science, artificial intelligence,engineering, electrical & electronic,operations research & management science