VASC Seminar Announcement
            =========================

Date: Friday, 12/01/00
Time: 11:00-12:00
Place: NSH 3002

Speaker: Rainer Stiefelhagen, University of Karlsruhe
Title:   Multimodal focus-of-attention tracking in meetings"

Visual cues play an important role in human communication. In this research 
I am interested in detecting and tracking at whom the participants of a 
meeting are looking at during the meeting. This could be useful for determining 
the addressee of speech acts, indexing a multimedia meeting, and analyzing 
participants' activity and attention during a meeting. In this talk, I will 
present my work on multimodal focus-of-attention tracking by fusing head 
pose and sound source information. We simultaneously track participants' 
faces using a panoramic camera. We then estimate head poses from facial 
images using neural networks. In addition, we detect who is speaking using 
microphones. Based on the information about the current speaker(s), we 
predict at whom participants are looking. Finally, the output of the audio- 
and video-based focus of attention are fused to obtain a combined 
audio-visual estimation for each person's focus of attention. I will 
discuss experimental results.