Using Symlet Decomposition Method, Fuzzy Integral and Fisherface Algorithm for Face Recognition Seyed Mohammad Seyedzade Department of Electrical Engineering, Iran Univ. of Science and Technology, Narmak, Tehran, Iran Email: sm.seyedzade@ieee.org Sattar Mirzakuchaki Department of Electrical Engineering, Iran Univ. of Science and Technology, Narmak, Tehran, Iran Email: m_kuchaki@iust.ac.ir Amir Tahmasbi Department of Electrical Engineering, Iran Univ. of Science and Technology, Narmak, Tehran, Iran Email: a.tahmasbi@ieee.org Abstract — In this paper, we have proposed an approach for face recognition by composing Symlet decomposition, Fisherface algorithm, and Sugeno and Choquet Fuzzy Integral. This approach consists of four main sections: the first section uses Symlet, one of the Wavelet families, to transform an image into four sub-images which are called approximate, horizontal, vertical and diagonal partial images respectively. The aim of this work is to extract intrinsic facial features. The second section of this approach uses Fisherface method which is composed of PCA and LDA. The reason for using this was that it is not sensitive to intensive light variations and facial expression and gesture. The third and forth section of this approach, are related to the aggregation of the individual classifiers by means of the fuzzy integral. Both Sugeno and Choquet fuzzy integral are considered as methods for classifier aggregation. In this paper, Olivetti Research Labs face database is used for acquiring experimental results. The approach presented in this paper, will lead to better classification performance compared to other classification methods. Keywords— Classifier aggregation, symlet decomposition, face database, face recognition, Fisherface method, fuzzy integral. I. INTRODUCTION In recent years different approaches for face recognition have been introduced. Extrinsic imaging parameters such as pose, illumination and facial expression still cause much difficulty in accurate face recognition. However, PCA and LDA methods are two common approaches for face recognition. PCA is a popular approach in image processing and communication theory that is quite often referred to as a Karhunen–Loeve (KL) transformation. The PCA approach exhibits optimality when it comes to dimensionality reduction. However, it is not ideal for classification purposes since it is too sensitive to undesirable light intensity variations and facial expression. To overcome this problem, Fisherface method, or Fisher’s linear discriminant (FLD), linear discriminant analysis (LDA) have been proposed [1]. Linear Discriminant Analysis, or simply LDA, is a well- known feature extraction technique that has been used successfully in many statistical pattern recognition problems. This technique maximizes the ratio of the determinant of between-class scatter matrix and within- class scatter matrix. In general, this method is used in conjunction with the PCA where the PCA technique first projects the set of images to a lower-dimensional space so that the resulting within-class scatter matrix, to be used by the FLD, becomes non-singular. FLD is capable of forming well- separated classes in a low-dimensional subspace, even under severe variation in lighting and facial expressions. Different approaches are used for extracting intrinsic facial features of image. One of the most powerful approaches for extracting intrinsic facial features is Wavelet transformation, which is capable of analysing multi-resolution decomposition; indeed, wavelet decomposition technique is used for extracting intrinsic facial features. The other authors only have used one approximate image among four sub-images relative to Wavelet decomposition. For instance, Chien performed face recognition using discriminant wavelet faces and nearest feature classifiers [2]. The initial incentive for this selection was that this approximate image has the best approximation of the main image in a space with smaller dimensions, and in fact it has the most energy among all the other four images. Sergent found that the low- frequency band and high-frequency components band performed different roles in the classification task. The low-frequency components contribute to the global description, while the high-frequency components contribute to the finer details required in the identification task [3]. In this paper, three other sub-images are used together with Symlet decomposition approximate image for providing auxiliary information. These four images are used as input images in Fisherface. The fusion of the individual classifiers is realized through fuzzy integration with fuzzy integral [4]. The ability of the fuzzy integral to combine the results of multiple sources of information has been researched in various application areas such as pattern recognition. Chiang developed hybrid fuzzy-neural systems for handwritten word recognition using self- organizing feature map and Choquet fuzzy integral [5]. This paper is organized in the following manner. The second section describes Symlet decomposition and procedure for image decomposition. In third section, Fisherface method principles are introduced. In fourth section, Integral measurement theory and Fuzzy Integral are explained. Section V describes the technique of face recognition realized by means of fuzzy integral and Fisherface algorithm and symlet decomposition. Simulation results in ORL database[6] will be explained 2010 Second International Conference on Computer Engineering and Applications 978-0-7695-3982-9/10 $26.00 © 2010 IEEE DOI 10.1109/ICCEA.2010.173 83 Authorized licensed use limited to: Iran Univ of Science and Tech. Downloaded on April 25,2010 at 20:00:49 UTC from IEEE Xplore. Restrictions apply.