Age Estimation Using Support Vector Machine

  • Ayad R. Abbas Computer Science Department, University of Technology, Baghdad, Iraq
  • Asaad R. Kareem Computer Science Department, University of Technology, Baghdad, Iraq
Keywords: Viola jones, linear discriminate analysis, machine earning, SVM classifier, Age estimation


Recently there has been an urgent need to identify the ages from their personal pictures and to be used in the field of security of personal and biometric, interaction between human and computer, security of information, law enforcement. However, in spite of advances in age estimation, it stills a difficult problem. This is because the face old age process is determined not only by radical factors, e.g. genetic factors, but also by external factors, e.g. lifestyle, expression, and environment. This paper utilized machine learning technique to intelligent age estimation from facial images using support vector machine (SVM) on FG_NET dataset. The proposed work consists of three phases: the first phase is image preprocessing include four stages: grayscale image stage, histogram equalization stage, face detection stage has been carried out using viola jones algorithm, it comprises for four steps namely: Haar like Feature, integral image, Adaboost training, and cascading classifier, the last stage of image preprocessing phase is cropping and resize stage. The second phase is data mining include two stages: feature extraction stage using linear discriminate analysis and machine learning stage using support vector machine. The last phase is age estimation and evaluation. The FG-net dataset is used which divided into seven classes in order to has been became increased accuracy and reduce the execution time, the first class represents 3-7 years, the second class represents 8-13 years, the third class represents 14-19 years, the forth class represents 20-25 years, the fifth class represents 26-30 years, the sixth class represents 31-40 years and the seven class represents 41-50 years. Then, the seven classes are combined into three classes depending on the number of features. The Experimental results display that the proposed system can grant high accuracy. The practical evaluation of the proposed system gives accuracy is 84%.

How to Cite
Abbas, A. R., & Kareem, A. R. (2018). Age Estimation Using Support Vector Machine. Iraqi Journal of Science, 59(3C), 1746-1756. Retrieved from
Computer Science