اطلاعات کوهنوردی الهام بخش طبقه بندی کنندگان برای تشخیص چهره

Facial recognition is a challenging issue in pattern recognition arising from the need for high security systems capable of overcoming the variability of the acquisition environment such as illumination, pose or facial expression. A broad range of recognition methods have been suggested, yet most are still unable to yield optimal accuracy. More recently, new methods based on swarm intelligence or classifiers combination have been devised in the field of facial recognition. Swarm intelligence based methods aim to achieve effective recognition accuracy by exploiting their global optimization capability. The combination of classifiers is a new trend allowing cooperation between multiple classifiers. In this work, two classifiers inspired from swarm intelligence are proposed: a bees algorithm based classifier and a decision tree based binary particle swarm optimization classifier. The two are then combined with a decision tree based fuzzy support vector machine by using the majority vote as an attempt to compensate for the weakness of single classifiers. Moreover, the impact of different characteristic features and space reduction methods has been examined namely, the Gabor magnitude and the Gabor phase congruency features in combination with PCA, LDA or KFA reduction space methods. The experiments were conducted on four popular databases: ORL, YALE, FERET and UMIST. The results revealed that the proposed swarm intelligence based classifiers are very effective compared to similar classifiers in terms of recognition accuracy.