سال انتشار: ۱۳۹۱

محل انتشار: بیستمین کنفرانس مهندسی برق ایران

تعداد صفحات: ۵

نویسنده(ها):

Kian Hamedani – Amirkabir University of Technology
Seyyed Ali Seyyed Salehi – Amirkabir University of Technology

چکیده:

This paper proposes a method using manifold learning by neural networks for identifying people while they are talking. In training phase people say the same sentence, we trainthe NN for learning low-dimensional nonlinear manifolds that are embedded in high-dimensional video space. After trainingphase we use another video of the same persons while they are saying another sentence for testing. Comparing the recognition results with other methods shows that our method outperformsother methods. Finally we achieve 98.4% of recognition rate