next up previous
Next: Head Tracking and Feature Up: FACEFACTS: Modeling Natural Facial Previous: FACEFACTS: Modeling Natural Facial

Data Collection

The data collection process should allow the subjects to move freely and naturally. The cameras and microphones should be placed in a way that is very unobtrusive, yet able to capture the changes in facial and vocal expressions. To fulfill this requirement, we designed and built a conversation space in our lab. The resources used to build the conversation space are : (i) Three Sony EVI-D30 pan-tilt cameras (ii) Two AKG 1000 cardioid microphones (ii) Quad splitter (iv) Hi8 video recorder (v) TV monitor (vi) Table and a set of chairs (vii) Cables, connectors, camera mounts etc.

Two of the cameras were directly behind and slightly above each subject to get a frontal view of the subjects participating in the conversation. To capture the entire scene, a wide angle lens was attached to the third camera and placed in the center corner. The outputs of the cameras were fed into a quad splitter. The quad splitter was connected to the recorder and the TV screen was used to monitor the recording process and to detect any problems. The whole system is almost invisible to the subjects and does not constrain them in any way.


  
Figure: Conversation Space Set-up
\begin{figure}
\center
\begin{tabular}[htb]{cc}
\epsfysize=1.0in
\epsfbox{/u...
...nversation scene &
(b) Input to the Hi8 recorder \\
\end{tabular}\end{figure}


next up previous
Next: Head Tracking and Feature Up: FACEFACTS: Modeling Natural Facial Previous: FACEFACTS: Modeling Natural Facial
Tanzeem Choudhury
2000-01-21