T.: Multiple face recognition from omnidirectional video

Meetings are an integral part of business life. In previous work, we have developed a physical awareness system called CAMEO (Camera As-sisted Meeting Event Observer) to record and process audio/visual information of a meeting. A very important task in meeting understand-ing is to know who is attending to the meet-ing and CAMEO’s task is to infer people’s iden-tity from video. In this paper, we present an approach to identify people from an omnidirec-tional video sequence. Two main novelties are proposed: first a new dimensionality reduction technique MODA (Multimodal Oriented Discrim-inant Analysis) is used to perform fast matching and second we show that using multiple spatio-temporal constraints the recognition performance greatly improves. The effectiveness and robust-ness of the proposed system is demonstrated over several real time experiments and a large data set of videos.