Speed dependencies of human gesture recognition

Abstract

Gestures are important communication medium. However, their semantic ambiguities make it difficult for computer systems to accurately recognize and deliver them. The interpretation of gestural motions is not even consistent in human-to-human communications. In this paper, we compare gestures in order to establish a method to reveal the influence of gesture speed on semantic interpretation. We captured whole-body motions as movie files with a depth camera and converted them into biological motion movie files. The speed of these motion images was then systematically altered, and a perceptual experiment was conducted on 11 participants. Using the results from the experiment, we identified speed dependencies of human gesture recognition.