Audiovisual Affective Design of Humanoid Robot Appearance and Voice Based on Kansei Engineering

Humanoid robots, characterized by their anthropomorphic design, have become increasingly common in various service areas. Nevertheless, the majority of current affective designs of humanoid robots primarily concentrate on the physical appearance while overlooking its (audiovisual) integration with voice. In this study, we propose simultaneously designing the appearance and voice of humanoid robots using Kansei Engineering, an effective method for optimizing the affective design of products. We first selected representative humanoid robots with different appearances and voices and constructed kansei space to capture users’ affective needs for these robots. Then, we decomposed appearances and parameterized voices to extract design features and orthogonalized these design features to generate prototypes. After that, we conducted an evaluation experiment to acquire users’ affective evaluations on the combinations of appearance and voice. Based on the data, relationship models between design features and users’ kansei images and holistic preferences were constructed using the back-propagation neural network. Furthermore, optimization design models were formulated and resolved through the genetic algorithm. Also, we conducted a validation experiment, and the results demonstrated that the optimized design schemes look harmonious in appearance, sound warmth in voice, and achieve a high level of audiovisual compatibility. The results suggest that the proposed approach can effectively optimize the audiovisual affective design of humanoid robot appearance and voice. Moreover, it can not only provide methodological support for the affective design of robots and other voice-based smart products but can also help to improve the affective experience quality and facilitate the application of robots in service areas.