Real-time Motion-based Gesture Recognition Using the GPU.

Abstract

In this paper we describe a real-time system for gesture recognition. Given an input video, we derive a set of motion features based on smoothed optical flow estimates. A user-centric representation of these features is obtained using face detection, and an efficient classifier is learned to discriminate between gestures. We develop a real-time system using GPU programming for implementing the classifier. We provide experimental results demonstrating the speed and efficacy of our system. 1