Gender Classification And Age Estimation Based On Face Images

Human face is all important bio-feature.It's complicated in structure and detail,yet it contains a quantity of information,such as gender and age.The human face recognition is well developed,while the predictions of face properties,as age and gender,are not yet well studied.This work designs an approach for gender classification and age estimation of human face based on Triangular norm,relevance projection,Fourier transform and support vector machine(SVM).The main content of this paper are as follows:No.1.Firstly,the image preprocessing is scaled and the skin pixels are extracted from the background based on the Gaussian Mixture Model and the thresholds are obtained based on the image information,with the threshold from the background of the skin area.Then,we introduce the triangular norm substitution product into the edge detection algorithm based on the idea of universal gravitation.And the gravitational force generated by the edge pixels is calculated.Finally,the vector is standardized,and an edge fuzzy set which can represent the membership degree of each position is established,and according to optimizing the brightness dependence,the edge images are compared by the comparison of binary images.On the FERET data set,the algorithm is compared with the traditional methods such as the Canny operator and the Sobel operator based on the triangle norm.The experimental results show that the proposed algorithm is superior to the traditional method in performance.No.2.In order to improve the performance of the proposed method,a new method based on supervised learning is proposed.This method can be applied to different facial analysis tasks.Firstly,the algorithm reduces the dimension of the face based on the weighted PCA algorithm and facial features are extracted;then,the algorithm is optimized and the error function of correlation projection is calculated;finally,the correlation error function in facial projection is minimized and theEuclidean distance of the feature vector is calculated for face gender classification.The proposed method was compared with other feature extraction methods,and the experiments were carried out on the FERET database.The results of experiment prove the method is efficient and obtain higher recognition accuracy than traditional methods.No.3.To solve this problem,we propose a feature extraction method based on Fourier-Mellin transform in the frequency domain invariants and the moment domain invariants.Firstly,the image brightness will be normalized to improve illumination variation.Then,the two algorithms are mixed to extract features,according to the Analytical Fourier-Mellin Transform(AFMT)invariants and the Orthogonal FourierMellin Moment(OFMM)invariants.Finally,the hybrid algorithm is classified and merged by the Nearest Neighbor Classifier(NNC)and the Correlation Coefficient Method(CCM).Lots of experiments on YALE and ORL face databases,the results show that the performance of our proposed method is superior than that of the traditional face recognition algorithms.No.4.In this paper,we find that in addition to appearance information,facial dynamics can also be exploited in age estimation.We propose a method to extract dynamic features for age estimation by using a person's smile.First,we evaluate the proposed system accuracy when using only dynamic facial,or separately for each facial region,or in the smile together.We compare these results with a combination of appearance and kinetics.We then used combinatorial features to test the effects of gender and expression spontaneity on system accuracy.Finally,the appearance characteristics IEF,GEF,BIF and LBP were calculated to evaluate the age estimation results by experiment complexity.Experiments on FERET face database show that our proposed method outperforms traditional algorithms.