| dc.description.abstract | Biometric system provides automatic identification of an individual based on a unique feature or characteristic possessed by the individual. Data used for this research are obtained from CASIA dataset consisting of 600 eye images from 30 people with 10 left eye images and 10 right eye images from each person. The data are divided into two categories, training data (360 images) and testing data (240 images). Each image contains eyelid, eyelash, sclera, iris, and pupil. The system uses an automatic segmentation based on a threshold to localize the iris collarette. A rubber sheet model is then used to normalize the image into a 20×240 matrix. Feature extraction is conducted using Haar wavelet transformation at level 0 (original image), level 1, level 2, level 3, and level 4. The features are fed into a backpropagation neural network for classification. Parameters observed in this research are number of hidden neurons, learning rate, and momentum at each level wavelet decomposition. Maximum generalization and training time are used as the performance metrics. The best parameters obtained at each level wavelet decomposition are used in testing using 5-fold cross validation. The best recognition of testing data considering the maximum generalization and the training time was obtained from wavelet decomposition level 3. | en |
