אירועים והרצאות בפקולטה למדעי המחשב ע"ש הנרי ומרילין טאוב

Multi-camera environments allow exploiting spatial redundancy that might be used by analysis algorithms to produce more robust results. These algorithms may be applied for a variety of tasks ranging from human 3D scene volume reconstruction, human motion capture, gesture analysis or focus of attention tracking. Combining this multi-camera information with audio information would allow defining multimodal data analysis systems. This talk will present two research lines within this scope. First, a Particle Filtering based algorithm towards human body motion capture is presented as the building block for a gesture recognition system. The output of this system is fed into a multimodal dance analysis system. Secondly, a multi-view head orientation algorithm is introduced as a technology to asses focus of attention in meetings. Audio information is also added to build up a multimodal approach to this problem. Results are presented showing the effectiveness of the proposed algorithms.