Human age estimation based on face images

Abstract

Nowadays, there are forensic artists who are able to draw out images of people and at the same time make realistic age progression. There are Swedish alcohol salespersons that have a professional skill in age recognition. But the question is can the computer perform the same function like the group of people mentioned? In this project, the student had developed a program for the computer to recognize the age group of the user via a webcam. The student made used of Haar-like features from OpenCV for face detection via a webcam. Upon the detection of face, the image was cropped and thereafter the features of the face will be extracted. Feature extraction was done using Gabor filters and local binary pattern. After the extraction of features, the program will classify the result into 4 categories. They are “Children”, “Teens”, “Adults” and “Senior Adults”. In this project, the student had trained adaboost and sets of results were obtained. The results obtained from Adabooost were thereafter applied to Error Correction Output Coding for the final prediction of the age group. The purpose of Error Correction Output Coding was to convert multi class problem to a two-class problem.