@phdthesis{digilib10894,
           month = {August},
           title = {PENGENALAN WAJAH MENGGUNAKAN JST  
BERDASARKAN EKSTRAKSI CIRI PCA DAN LDA
},
          school = {PERPUSTAKAAN UIN SUNAN KALIJAGA},
          author = {NIM. 07650009 MUHAMMAD FAZLUR RAHMAN  },
            year = {2012},
            note = {Shofwatul ?Uyun, M.Kom},
        keywords = {face recognition system, artificial neural network, back propagation,
PCA, LDA, image
},
             url = {https://digilib.uin-suka.ac.id/id/eprint/10894/},
        abstract = {Humans have the intelligence of a very complex multi-intelligence. Just as 
human is able ti quickly recognize, memorize and distinguish each person?s face
with an expression of previously known and different circumstances. Unlike
human, computer system needs to be trained to pick the skills that are similar to
humans. Biometric is the ability of the verification and identification based on
physical characteristic and human behavior. Face recognition system is an
example of biometric based on physical characteristic that can help computer
system to have the similar capability to humans intelligence. 
This research uses 14 facial images as the trained data that?s extracted to
be covarian matrics then the eigen values taken from each datum of images using
method of principal component analysis (PCA) and linear discriminant analysis
(LDA). Each datum produce 4 eigen values which become the inputs to the
training algorithm of artificial neural network of back propagation and the
application algorithm of back propagation. The training algorithm earns outputs in
form of optimal weight values which become reference to the application
algorithm. 
The system is capable of extracting facial images into eigen values using
PCA and LDA. Face recognition system is also capable of doing training data
using an artificial neural network of back propagation. The system can also
recognize facial images with successfull procentage 77,77\% by structure of
artificial neural network of back propagation: 4 nodes at input layer, 8 nodes at
hidden layer and 3 nodes at output layer using epoch value of training 60x10
4
}
}