An automatic training process using clustering algorithms for face recognition system

The paper deals with automatic training process using clustering algorithms. We provide a comparative study of several clustering algorithms such as K-means, self-organizing map (SOM) and DBSCAN. We use these algorithms for automatic selection of appropriate training samples for face recognition system. We provide an overview of selected methods and we compare their performance on the CMU PIE face database with different types of preprocessing. Based on results published in this paper we propose a part of face recognition system for the Next-Generation Hybrid Broadcast Broadband (HBB-Next) project.