Gait human identification, considered as new biometric number, became a very active research field in the last few years. Realized works, until now, was based on classical techniques. In our work, we tried to extract the best characteristics that verify the uniqueness of the silhouette of the individual and the classification was done by two methods: multi-class Separator Vast Marge (SVM) and neural networks radial basis (RBF). The recognition rates obtained are satisfactory regarding the two methods (Highest 92.12% for SVM and 90.77% for RBF). By comparing the two methods we note that the main difference remains in the time that is expensive in SVM (one against one) compared to the RBF, which are known for their speed, it allows us to say that the RBF are more appropriate for real-time applications.
-
Votre commentaire
Votre commentaire s'affichera sur cette page après validation par l'administrateur.
Ceci n'est en aucun cas un formulaire à l'adresse du sujet évoqué,
mais juste un espace d'opinion et d'échange d'idées dans le respect.
Posté Le : 28/09/2022
Posté par : einstein
Ecrit par : - Ouddane Wafaa - Benyettou Mohamed
Source : Models & Optimisation and Mathematical Analysis Journal Volume 1, Numéro 1, Pages 33-39 2012-12-17