PENGENALAN EKSPRESI WAJAH MENGGUNAKAN WAVELET DAN BACKPROPAGATION

SAPUTRO, IMMANUELA PUSPASARI (2015) PENGENALAN EKSPRESI WAJAH MENGGUNAKAN WAVELET DAN BACKPROPAGATION. S2 thesis, UAJY.

	Text (Halaman Judul) MTF002090.pdf Download (637kB)
	Text (Bab I) MTF102090.pdf Download (152kB)
	Text (Bab II) MTF202090.pdf Download (634kB)
	Text (Bab III) MTF302090.pdf Restricted to Registered users only Download (210kB)
	Text (Bab IV) MTF402090.pdf Restricted to Registered users only Download (1MB)
	Text (Bab V) MTF502090.pdf Download (1MB)

Abstract

Facial expression recognition has been an interested topic research for a recent decade. Many methods are proposed for the construction of automatic facial expression systems. An image is spatial dimension image that contains color information. This study aimed to test backpropagation network accuracy in the classification of person's facial expressions. For pre processing, combined of Otsu thresholding and Adaptive thresholding method are used. In order to recognize facial expressions images, a set of Facial expression needs to be extracted and Haar wavelet and Gabor wavelet are used to get that feature extraction. For classification, Nguyen Widrow weights initialization, sigmoid biner function, gradient descent train function, learngd learning function are used in backpropagation network. The architecture using two hidden layers and three neuron for output layer. Value target of Mean Square Error (MSE) is 0,001. The network will be trained by 70 data consisting of 10 people and each person has seven expressions. For testing purpose, the network wiil test using other data that has not been previously trained, composed of five people with seven expressions. Based on test results, the vector input image obtained by Haar wavelet transformation has accuracy in classification of facial expression by 85.19% and 83.60% on Gabor wavelet.

Actions (login required)

View Item