Saliency-Driven Real-Time Video-to-Tactile Translation

Abstract

Tactile feedback coordinated with visual stimuli has proven its worth in mediating immersive multimodal experiences, yet its authoring has relied on content artists. This article presents a fully automated framework of generating tactile cues from streaming images to provide synchronized visuotactile stimuli in real time. The spatiotemporal features of video images are analyzed on the basis of visual saliency and then mapped into the tactile cues that are rendered on tactors installed on a chair. We also conducted two user experiments for performance evaluation. The first experiment investigated the effects of visuotactile rendering against visual-only rendering, demonstrating that the visuotactile rendering improved the movie watching experience to be more interesting, immersive, and understandable. The second experiment was performed to compare the effectiveness of authoring methods and found that the automated authoring approach, used with care, can produce plausible tactile effects similar in quality to manual authoring.