Head gestures recognition

Abstract

In this paper we describe a system that automatically detects and recognizes human head gestures such as nodding and shaking in complex background conditions using a cheap Web Cam under uncontrolled conditions. The images of the head, captured at 20 frames per second, are very noisy and are of a low resolution. In the proposed system, the invariant moments of each image captured is extracted and is fed into a recognition system that uses discrete Hidden Markov Models (HMMs) to classify the head gestures. The system achieves an average success rate of 87%. The system can successfully run on any low to high end PC connected to a USB Web Cam without any manual initialization.