Through body centered cameras and machine vision techniques, several examples of contextually aware interfaces are presented. By observing context, the computer can aid in task and interruption management, provide just-in-time information, and make helpful predictions of future behavior. While larger annotated data sets are necessary to test the techniques used for the Patrol task, the preliminary results are promising. Additional methods such as optical flow or motion differencing may be added to determine if the user is standing, walking, running, visually scanning the scene, or using the stairs. By using the new apparatus to analyze video and audio from two simultaneous participants, player interaction might be modeled. Hopefully, with development, such a system will be used to observe and model everyday user tasks and human to human interactions as well.