Gender Classification from Facial Images Using Combinations of Wavelet, PCA/LDA AND SVM

Recognition gender from face are very important in todays security system. It simplifies the task of face analysis and recognition. This article presents two ways to classify gender using facial expressions. The first method combines PCA with wavelets while the second method combines LDA with wavelets for feature extraction. Embedded wavelet decomposition overcomes the limitations of PCA and LDA and improves gender classification. First, use wavelet to extract the predicted face image from the original face image. This wavelet characteristic is robust to local effects caused by changes in illumination, direction, and noise. Then, the predicted image is projected into low-dimensional space using the PCA/LDA technique to find the disparity vectors. Then the wavelet-based PCA/LDA feature vector is applied to the SVM/KNN classifier for the image to extract important information (image) and remove irrelevant information from the face sale. PCA/LDA is applied to these predicted images to find the difference vectors. The distribution value of the extraction method results was analyzed using kNN and SVM classifiers. To compare the proposed method, experiments were conducted on facial images taken in cold and unconditioned conditions. The suggested work in this paper gives acceptable output both image data taken on a natural background and on the data of images taken on an uncontrolled background. Wavelet decomposition using PCA and RBF SVM gives classification rate more than 95% with small feature vectors on Real, FERET, SUMS, FEI A and FEI B databases