next up previous
Next: Scenarios Up: System Overview Previous: Object Recognition System

   
Hardware

Currently, the system is fully tetherless with wireless radio connections allowing the user to roam around a significant amount of space (i.e. a few office rooms). Plans for evolving the system into a fully self-sufficient, compact and affordable form are underway. More powerful video processing in commercial units such as the PC104 platform or the VIA platform would eventually facilitate this process. However, for initial prototyping, a wireless system linked to off board processing was acceptable.


  
Figure: The Wearable Hardware System
\begin{figure}\center
\begin{tabular}[b]{c}
\epsfxsize=3.1in
\epsfbox{fig-uist-98/wearable2.eps}
\end{tabular}\end{figure}

Figure 5 depicts the major components of DyPERS which are worn by the user during operation. The user dons a Sony GlassTron heads-up display with a semi-transparent visor and headphones. Attached to the visor is an ELMO video camera (with wide angle lens) which is aligned as closely as possible with the user's line of sight [Starner et al. , 1997]. Thus the vision system is directed by the user's head motions to interesting objects. In addition, a nearby microphone is incorporated. The A/V data captured by the camera and microphone is continuously broadcast using a wireless radio transmitter. This wireless transmission connects the user and the wearable system to an SGI O2 workstation where the vision and other aspects of the system operate. The workstation collects the A/V data into clips, scans the visual scene using the object recognition system, and transmits the appropriate A/V clips back to the user. The clips are the rendered as an overlay via the user's GlassTron. Two A/V wireless channels are used at all times for a bidirectional real-time connection (user to SGI and SGI to user) [Mann, 1996].


next up previous
Next: Scenarios Up: System Overview Previous: Object Recognition System
Tony Jebara
1998-10-07