JIITA, vol. 1, no. 3, pp.34-39, Oct, 2017
DOI: 10.22664/ISITA.2017.1.3.34
Facial Expression Recognition using Region-SIFT
Hyunduk Kim, Sang-Heon Lee, Myoung-Kyu Sohn, Byunghun Hwang, Hyunseok Choi
Department of IoT and Robotics Convergence Research, DGIST, Daegu, Korea
Abstract: Facial expression recognition is a challenging field in numerous re-searches, and impacts important applications in many areas such as human–computer interaction and data-driven animation, etc. In this paper, we propose facial expression recognition system, which consists of three process. First step is face alignment using active shape model (ASM). Second step is feature extraction step. Because of high recogni-tion rate, Scale Invariant Feature Transform (SIFT) feature extraction method is widely applied to many recognition problems. However, it takes a long processing time. In order to solve the problem, we intro-duce Region-SIFT feature extraction method based on facial landmark. In this process, we define the sub-region of face, such as eye, nose, and mouth. And then we extract SIFT feature from each sub-region. In or-der to reduce dimension of feature we employ Principal Component Analysis (PCA) and Linear Discriminant Analysis (LDA). Finally, it is connected to non-linear Support Vector Machines (SVM) for efficient classification. The performance evaluation of proposed method was performed with the CK facial expression database and JAFFE data-base. The experimental results demonstrate that our proposed ap-proach is very efficient to classify the facial expression. As a result, the method using Region-SIFT showed performance improvements of 14.52% and 19.75% compared to previous method using SIFT for CK database and JAFFE database, respectively.
Keyword: Facial expression recognition; Face alignment; Local descriptor; Scale Invariant Feature Transform
Full Paper: Full Paper: