Back close

Emotion recognition from facial expressions of 4D videos using curves and surface normals

Publication Type : Conference Proceedings

Publisher : Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics

Source : Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), Springer Verlag, Volume 10127 LNCS, p.51-64 (2017)

Url : https://www.scopus.com/inward/record.uri?eid=2-s2.0-85011325162&doi=10.1007%2f978-3-319-52503-7_5&partnerID=40&md5=90567b775736cf8aa0a53d7bcb30bf14

ISBN : 9783319525020

Keywords : Classifiers, Deformation, Deformation matrix, Emotion recognition, Face recognition, Feature vectors, Human computer interaction, Image retrieval, Matrix algebra, Parallel curves, Radial curves, Speech recognition, Support vector machines, Surface normals, Vectors

Campus : Bengaluru

School : Department of Computer Science and Engineering, School of Engineering

Department : Computer Science

Year : 2017

Abstract : In this paper, we propose and compare three methods for recognizing emotions from facial expressions using 4D videos. In the first two methods, the 3D faces are re-sampled by using curves to extract the feature information. Two different methods are presented to resample the faces in an intelligent way using parallel curves and radial curves. The movement of the face is measured through these curves using two frames: neutral and peak frame. The deformation matrix is formed by computing the distance point to point on the corresponding curves of the neutral frame and peak frame. This matrix is used to create the feature vector that will be used for classification using Support Vector Machine (SVM). The third method proposed is to extract the feature information from the face by using surface normals. At every point on the frame, surface normals are extracted. The deformation matrix is formed by computing the Euclidean distances between the corresponding normals at a point on neutral and peak frames. This matrix is used to create the feature vector that will be used for classification of emotions using SVM. The proposed methods are analyzed and they showed improvement over existing literature. © Springer International Publishing AG 2017.

Cite this Research Publication : SaSai Prathusha, Dr. Suja P., Dr. Shikha Tripathi, and Louis, Rc, “Emotion recognition from facial expressions of 4D videos using curves and surface normals”, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 10127 LNCS, pp. 51-64, 2017.

Admissions Apply Now