Una Aproximación De Clasificación De Rostros Usando Antropometría Facial

Abstract

The face images processing is one of the study areas within the field of computer vision. Thus there are works mainly in face recognition, emotions identification, among others, however the development of automatic systems of age estimation is a challenge that is still under investigation. In the present work, a procedure to classify a face in a determined age range is proposed. For said proposal, theories of craniofacial growth and facial anthropometry are analyzed, deriving in a selection of anthropometric parameters that represent discriminant characteristics for the distinction of faces at different ages. These parameters were used to generate a classification model on Weka platform using the SVM, Knn, Naïve Bayes and C4.5 algorithms. Cross validation at 10 folds was used for each algorithm. The highest accuracy was obtained with Knn with 7 neighbors and it was 75. 28%. This verifying the usefulness of the anthropometric distances selected for the recognition of age in face images.