Finger track — A robust and real-time gesture interface

Abstract

Real-time computer vision combined with robust gesture recognition provides a natural alternative to traditional computer interfaces. Human users have plenty of experience with actions and the manipulation of objects requiring finger movement. In place of a mouse, users could use their hands to select and manipulate data. This paper presents a first step in this approach using a finger as a pointing and selection device.

A major feature of a successful tracking system is robustness. The system must be able to acquire tracked features upon startup, and reacquire them if lost during tracking. Reacquisition should be fast and accurate (i.e. it should pick up the correct feature). Intelligent search algorithms are needed for speedy, accurate acquisition of lost features with the frame. The prototype interface presented in this paper is based on finger tracking as a means of input to applications. The focus of the discussion is how the system can be made to perform robustly in real-time. Dynamically distributed search windows are defined for searching within the frame. The location and number of search windows are dependent on the confidence in the tracking of features. Experimental results showing the effectiveness of these techniques are presented.