Audiovisual Person AuthenticationProfessor Michael Wagner Tuesday 7th June 2005 at 11am AbstractResearch over the past two decades in speaker recognition and in face recognition has recently generated viable commercial products in both areas. Both technologies, however, while effective and useful in specific environments, do not presently adapt well to ecological changes, such as variations of illumination, room acoustics or channel characteristics. Moreover, both technologies can be vulnerable to replay attack. Person authentication by a combination of face and voice features offers the advantages of higher reliability and robustness owing to the use of two largely independent feature sets, as well as a higher level of confidence in the "liveness" of the synchronous dynamic audiovisual signal. This talk gives an overview of existing face recognition and speaker recognition methods and presents recent work on person authentication and liveness verification using three different audiovisual data corpora. Short resumeMichael Wagner obtained a Diplomphysiker degree from the University of Munich in 1973 and a PhD in computer science from the Australian National University in 1979. He has held research and teaching positions at Technical University of Munich, University of Duisburg, National University of Singapore, University of Wollongong, Australian Defence Force Academy, ANU and Nixdorf Corporation. From 1993 to 1996 he was project leader of the TRUST multimedia authentication project at ANU and since 1996 he has been professor of computing at the University of Canberra. His publications are mainly in the areas of speech and speaker recognition. Michael Wagner was the foundation president of the Australian Speech Science and Technology Association (ASSTA). He chairs ASSTA's National Spoken Language Database Committee and he is a Board member of the International Speech Communication Association (ISCA). |