Yasutoshi Makino

Designing Vibrotactile Sensations Using Machine Learning

Vibrotactile feedback plays a key role in enhancing the user experience in video games and similar applications. When generating this feedback, haptic sensations must sometimes be created for imaginary or unknown events. Therefore, it is desirable to be able to freely design the intended tactile experience. In this study, we introduce approaches to designing desired vibrotactile sensations using machine learning. First, we present a study investigating the minimum number of parameters required to design vibrotactile feedback. This study aims to reduce the search space to an appropriate scale for optimizing parameters through machine learning. The second part describes a method that constructs a minimally sufficient latent vector within the GAN framework and optimizes it using a human-in-the-loop model. We demonstrate that the system can generate vibrations corresponding to several predefined categories.