Head pose estimation by regression algorithm

Head pose estimation is a very in-depth topic in the context of biometric recognition, especially in video surveillance, because the rotation of the head can affect the recognition of some features of the face. Being able to recognize in advance the pose of the head in pitch, yaw and roll enable frontalization or the extraction of a frame in which a face is frontal in order to allow a more accurate recognition. In this work the Web-Shaped Model algorithm is used for a coding of the pose of the face and then we apply regression algorithms to predict the pose of the face. The proposed approach stimulates the sensitivity of the regression methods to identify the head pose estimation. The goals is to predict the value of the dependent variable for the three angular values, for which some information relating to the explanatory variables is available, in order to estimate the effect on the dependent variable. The presented method is tested on some of the most well-known datasets for the head pose estimation as Biwi, AFLW2000 and Pointing'04 and compared with the various state of the art methods that use these datasets.