Estimation of age groups using facial recognition features

Facial recognition has been used to detect emotions, gender, expressions and identity. These topics have been extensively studied. But, automatic Age Estimation is a topic that has not been researched much. The basis of our topic is that there are features on the human face that change as our age increases and in our research we utilize these features for age group classification. We classify the images into 6 age groups- (0-6, 8-12, 15-20, 25-32, 38-53, 60-100). The process involves 3 stages: -Pre-processing, Feature Extraction, Classification. Preprocessing includes commotion decreasing, standardization and change of the crude information into structure that is suitable for the pattern recognition. A small set of good features is selected from the available features, in order to provide best matching information. This process discovers important features to obtain an effective and improved answer for a given problem which is the next step. At last the classification or clustering stage is performed. Classification comprises of allotting a class mark to a given pattern while clustering discoveries homogeneous small groups in information. We also deal with comparison study of various feature extraction techniques (Haar feature extraction, HOG feature extraction) and classification techniques (Naive Bayes, SVM, KNN, Neural Network-Back propagation algorithm) and finally try to determine the best possible combination for age group classification. Convolutional Neural Networks is also applied to estimate the age groups.