Machine Vision and Gesture Recognition

Gesture recognition allows the computer to understand human’s body language, and therefore can be useful in many applications for natural human computer interaction. The use of gestures for interaction is especially useful in a virtual reality environment. Recent development in display technologies, like large-scale displays and table-top displays, also opens the realm for developing new gestural interaction techniques. Gesture recognition also has application in robotic control and teleconference. Although there are special devices, such as Cybergloves, that can be worn to achieve the goal for gesture recognition, such devices are often expensive and unwieldy [4]. Cameras, especially web cameras, are getting cheaper and cheaper. As a result, since 1990s, there has been an increasing amount of research in computer vision to provide a cheaper and more convenient solution for gesture recognition. Hand gesture is the most expressive part of the body language, and it also presents many challenging computer vision problems due to its complexity. Gestures can be classified into two categories: temporal gestures and static hand postures. This paper reviews some of the vision-based gesture recognition techniques for both categories. Both the merits and limitations of the techniques are discussed.