Abstract: There are many demonstrations of real-time interactions with virtual characters in 3D environments in which expensive equipment is affixed to the user, or a specialized environment is set up using known backgrounds and high speed infrared camera systems. These system are very reliable for tracking very detailed movements of the user. However, for many situations this precision is not needed. Several virtual reality applications and interactions only need estimations of the user's actions. This paper describes how an inexpensive stereo webcam may be used to approximate the upper body position of the user.