 |
Lecturer publishes research in the global magazine for classification of sex in the digital facial image depending on the inclination transformative |
Lecturer publishes research in the global magazine for classification of sex in the digital facial image depending on the inclination transformative
In this research provided a new proposal for the classification of sex based on the digital image of the face, and consists of the proposed classification algorithm of two phases: phase of training and testing phase.
IT included the training phase five steps for classifying sex images, whether male or female, in the first step is the face chopping of the image in order to get rid of the background of unwanted image, either reduce and minimize the noise by using the inclination conversion (Slantlet transform) is in the second step, after the implementation of the tendency transfer of the image is extracted characteristics of them using the principle of the analysis of the components (PCA), the third step in order to reduce the number of dimensions without loss of information, (where the intrinsic value (Eigen value) are used as a guide characteristics), the last step is highlighted in the decision whether it's male or female is applying machine support vector (SVM) on characteristic routers. An experimental result for the classification of sex has achieved accuracy 89 percent when the dependence on wavelength conversion (Wavelet), and either the implementation of the classification on the basis of the tendency of transfer (Slant let) the accuracy of the results has reached 93% using the same number of test.