B. T. Taylor, V. Michael Bove, Jr.

Abstract

We propose a vision of a grasp-based interaction system where users' intentions are inferred by the way they hold and interact with a device. In this paper we specifically discuss the Bar of Soap, a multi-function handheld prototype that uses grasp-based interactions to switch between modes. This prototype relies on the hypothesis that users share a set of stereotyped grasps associated with common multi-function modes. We show that using common machine learning techniques our device can reliably distinguish five separate classes based on the users’ grasps. While this interaction is currently implemented in a multi-function handheld, we anticipate the existence of many scenarios where grasp recognition could provide a more intuitive or useful interface.