A problem that is commonly faced by large-scale population system is the high-dimensionality of data that needs to be processed at a given time. In this paper, a new face recognition training structure is proposed in which the large-scale population is split into smaller groups to be processed separately. To improve classification the proposed system uses global and local linear discriminant analysis together with a similarity measure to maximize the separation of features within each group. Implementations of the proposed structure indicate that the presented structure has a better performance and faster training time compared to a conventional training structure.
|Title of host publication||Proceedings - 2010 3rd IEEE International Conference on Computer Science and Information Technology, ICCSIT 2010|
|Publisher||IEEE Computer Society|
|Number of pages||4|
|Publication status||Published - 01 Jan 2010|