Output details
11 - Computer Science and Informatics
Manchester Metropolitan University
Behavioral consistency extraction for face verification
<23>This reports a comprehensive study of the recognition of video sequences of faces, modelling the consistent patterns of expressive and communicative configurations via a sub-space method within an Appearance Model approach, minimizing the Description Length of each complete sequence. Comprehensive testing on the public BANCA database shows significantly improved recognition relative to comparable algorithms. Significantly, it shows that distinctive behavioural patterns can be detected across sequences of the same person, even if they are present in different magnitudes. The impact will be seen in the design of moving face recognition algorithms, and studies of facial expressive interpretation.