Automatic face recognition by support vector machines

Automatic face recognition, though being a hard problem, has a wide variety of applications. Support vector machine (SVM), to which model selection plays a key role, is a powerful technique for pattern recognition problems. Recently lots of researches have been done on face recognition by SVMs and satisfying results have been reported. However, as SVMs model selection details were not given, those results might have been overestimated. In this paper, we propose a general framework for investigating automatic face recognition by SVMs, with which different model selection algorithms as well as other important issues can be explored. Preliminary experimental results on the ORL face database show that, with the proposed hybrid model selection algorithm, appropriate SVMs models can be obtained with satisfying recognition performance.