Vol.7 (2017) No. 5 ISSN: 2088-5334 A Study on Facial Expression Recognition Using Local Binary Pattern Shahreen Kasim # , Rohayanti Hassan * , Nur Hadiana Zaini * , Asraful Syifaa’ Ahmad * , Azizul Azhar Ramli # , Rd Rohmat Saedudin + # Soft Computing and Data Mining Centre, Faculty of Computer Science and Information Technology, Universiti Tun Hussein Onn, Johor, Malaysia * Software Engineering Research Group, Faculty of Computing, Universiti Teknologi Malaysia, 81310 Skudai, Johor, Malaysia E-mail: rohayanti@utm.my + School of Industrial Engineering, Telkom University, 40257 Bandung, West Java, Indonesia Abstract— How to get the proper combination of feature extraction and classification is still crucial in facial expression recognition, and it has been addressed conducted over two decades. Hence, if inadequate features are used, even the best classifier could fail to achieve the accurate recognition. Therefore, Local Binary Pattern (LBP) is used as a feature extraction technique for facial expressions recognition where it is evaluated based on statistical local features. LBP is proven successful technique by the recent study due to its speed and discrimination performance aside of robust to low-resolution images. For the classification, Support Vector Machine is chosen, and the algorithm is implemented in MATLAB and tested on JAFFE (Japanese Female Facial Expressions) database in order to achieve the objectives and the goal of this research which is to obtain high accuracy in facial expressions and identify the seven basic facial expressions. The performance of feature extraction and classification is evaluated based on the recognition accuracy. The observation on results obtained in facial expressions recognition rate indicated the effectiveness of the proposed algorithm based on SVM-LBP features. Keywords— facial expression recognition; feature extraction; local binary pattern; support vector machine; JAFFE I. INTRODUCTION Biometric is not foreign things in the technology of computer science which has been adapted for various applications. Basically, biometric is an authentication of behavioural characteristics and physical of individuals as a form of access control or identification. The well-known biometric authentication such iris recognition, fingerprint, DNA, palm print, hand geometry, face and voice recognition has become widespread in world applications, yet there are still have challenges to overcome. As compared to fingerprints and iris, face recognition has diverse advantages as it non-contact process. A numerous study has been conducted over two decades to address the problem of face recognition, especially in facial expression. Even though a lot of approaches had been discussed, expression recognition is difficult task to achieve the optimal pre-processing, feature extraction or selection and classification under a certain condition. Nowadays, a study on the combination of face representation and classification is crucial in facial expression recognition. Moreover, the best classifier could fail to obtain accurate recognition if inadequate features are used. Numerous studies have been made on recognizing facial expression with a high accuracy yet remains difficult due to subtlety, complexity, and variability of facial expressions. Furthermore, low-resolution images in real- world environments make real-life expression recognition much more difficult [1]. It is necessary to extract important facial features for classifying facial expressions into variance categories which contribute in identifying proper and expression. Facial basic expressions, for example, are sad, happy, disgust, fear, surprise, angry and neutral. Those particular facial expressions of emotions are termed as universal emotion by Ekman [2] hence over a decade; other researchers have used a similar method in their research [3], [4], [5]. Local Binary Pattern (LBP) is proposed in this paper which originally for texture analysis and recently have been introduced to represent faces in facial images analysis. Zhang [5] has proposed an improved approach for facial expression analysis where LBP histogram of different block sizes of a face image is used as feature vectors and then various facial expressions were classified using Principal Component Analysis (PCA). Lajevardi et al. [6] have used LBP in their 1621