Biometric Bits 2006-01-No,03

A strategy of color image based human face representation is first proposed. Then, based on this representation, complex Eigenfaces technique is developed for facial feature extraction. Finally, we test our idea using the AR face database. The experimental result demonstrates that the proposed color image based complex Eigenfaces method is more robust to illumination variations than the traditional grayscale image based Eigenfaces. Contact Information Jian Yang Email: [email protected] URL: http://www4.comp.polyu.edu.hk/~biometrics/ Contact Information David Zhang Email: [email protected] URL: http://www4.comp.polyu.edu.hk/~biometrics/ Contact Information Yong Xu Email: [email protected] Contact Information Jing-yu Yang Email: [email protected] =================================== 11. Yunhong Wang, Yiding Wang, Anil K. Jain, Tieniu Tan: Face Verification Based on Bagging RBF Networks. 69-77 Electronic Edition (link) BibTeX Face Verification Based on Bagging RBF Networks Yunhong Wang1 Contact Information, Yiding Wang2 Contact Information, Anil K. Jain3 Contact Information and Tieniu Tan4 Contact Information (1) School of Computer Science and Engineering, Beihang University, Beijing, 100083, China (2) Graduate School, Chinese Academy of Sciences, Beijing, 100049, China (3) Department of Computer Science & Engineering, Michigan State University, East Lansing, MI 48824, (4) National Laboratory of Pattern Recognition, Institute of Automation, Chinese Academy of Sciences, P.O. Box 2728, Beijing 100080, P.R. China Abstract Face verification is useful in a variety of applications. A face verification system is vulnerable not only to variations in ambient lighting, facial expression and facial pose, but also to the effect of small sample size during the training phase. In this paper, we propose an approach to face verification based on Radial Basis Function (RBF) networks and bagging. The technique seeks to offset the effect of using a small sample size during the training phase. The RBF networks are trained using all available positive samples of a subject and a few randomly selected negative samples. Bagging is then applied to the outputs of these RBF-based classifiers. Theoretical analysis and experimental results show the validity of the proposed approach. file:///C|/Documents%20and%20Settings/Dorothea%20M....uments/My%20Webs/mysite11/BDG/BioBits2006-0103.html (20 of 83)1/6/2006 7:17:27 AM Biometric Bits 2006-01-No,03 Contact Information Yunhong Wang Email: [email protected] Contact Information Yiding Wang Email: [email protected] Contact Information Anil K. Jain Email: [email protected] Contact Information Tieniu Tan Email: [email protected] =================================== 12. Wangmeng Zuo, Kuanquan Wang, David Zhang: Improvement on Null Space LDA for Face Recognition: A Symmetry Consideration. 78-84 Electronic Edition (link) BibTeX Improvement on Null Space LDA for Face Recognition: A Symmetry Consideration Wangmeng Zuo1, Kuanquan Wang1 and David Zhang2 (1) School of Computer Science and Technology, Harbin Institute of Technology, Harbin, 150001, China (2) Biometrics Research Centre, Department of Computing, The Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong Abstract The approximate bilateral symmetry of human face has been explored to improve the recognition performance of some face recognition algorithms such as Linear Discriminant Analysis (LDA) and Direct-LDA (D-LDA). In this paper we summary the ways to generate virtual sample using facial symmetry, and investigate the three strategies of using facial symmetric information in the Null Space LDA (NLDA) framework. The results of our experiments indicate that, the use of facial symmetric information can further improve the recognition accuracy of conventional NLDA. =================================== 13. Cheng Zhong, Tieniu Tan, Chenghua Xu, Jiangwei Li: Automatic 3D Face Recognition Using Discriminant Common Vectors. 85-91 Electronic Edition (link) BibTeX Automatic 3D Face Recognition Using Discriminant Common Vectors Cheng Zhong1 Contact Information, Tieniu Tan1 Contact Information, Chenghua Xu1 Contact Information and Jiangwei Li1 Contact Information (1) National Laboratory of Pattern Recognition, Institute of Automation, Chinese Academy of Sciences, Beijing, 100080, P.R. China Abstract In this paper we propose a fully automatic scheme for 3D face recognition. In our scheme, the original 3D data is automatically converted into the normalized 3D data, then the discriminant common vector file:///C|/Documents%20and%20Settings/Dorothea%20M....uments/My%20Webs/mysite11/BDG/BioBits2006-0103.html (21 of 83)1/6/2006 7:17:27 AM Biometric Bits 2006-01-No,03 (DCV) is introduced for 3D face recognition. We also compare DCV with two common methods, i.e., principal component analysis (PCA) and linear discriminant analysis (LDA). Our experiments are based on the CASIA 3D Face Database, a challenging database with complex variations. The experimental results show that DCV is superior to the other two methods. Contact Information Cheng Zhong Email: [email protected] Contact Information Tieniu Tan Email: [email protected] Contact Information Chenghua Xu Email: [email protected] Contact Information Jiangwei Li Email: [email protected] =================================== 14. Xiao-Sheng Zhuang, Dao-Qing Dai, Pong Chi Yuen: Face Recognition by Inverse Fisher Discriminant Features. 92-98 Electronic Edition (link) BibTeX Face Recognition by Inverse Fisher Discriminant Features Xiao-Sheng Zhuang1, Dao-Qing Dai1 Contact Information and P.C. Yuen2 Contact Information (1) Center for Computer Vision and Department of Mathematics, Sun Yat-Sen(Zhongshan) University, Guangzhou 510275, China (2) Department of Computer Science, Hong Kong Baptist University, Hong Kong Abstract For face recognition task the PCA plus LDA technique is a famous two-phrase framework to deal with high dimensional space and singular cases. In this paper, we examine the theory of this framework: (1) LDA can still fail even after PCA procedure. (2) Some small principal components that might be essential for classification are thrown away after PCA step. (3) The null space of the within-class scatter matrix Sw contains discriminative information for classification. To eliminate these deficiencies of the PCA plus LDA method we thus develop a new framework by introducing an inverse Fisher criterion and adding a constrain in PCA procedure so that the singularity phenomenon will not occur. Experiment results suggest that this new approach works well. Contact Information Dao-Qing Dai Email: [email protected] Contact Information P.C. Yuen Email: [email protected] =================================== 15. Hwanjong Song, Ukil Yang, Sangyoun Lee, Kwanghoon Sohn: file:///C|/Documents%20and%20Settings/Dorothea%20M....uments/My%20Webs/mysite11/BDG/BioBits2006-0103.html (22 of 83)1/6/2006 7:17:27 AM Biometric Bits 2006-01-No,03 3D Face Recognition Based on Facial Shape Indexes with Dynamic Programming. 99-105 Electronic Edition (link) BibTeX 3D Face Recognition Based on Facial Shape Indexes with Dynamic Programming Hwanjong Song1 Contact Information, Ukil Yang1 Contact Information, Sangyoun Lee1 Contact Information and Kwanghoon Sohn1 Contact Information (1) Biometrics Engineering Research Center, Dept. of Electrical & Electronic Eng., Yonsei University, 134 Shinchon-dong, Seodaemun-gu, Seoul, 120-749, Korea Abstract This paper describes a 3D face recognition method using facial shape indexes. Given an unknown range image, we extract invariant facial features based on the facial geometry. We estimate the 3D head pose using the proposed error compensated SVD method. For face recognition method, we define and extract facial shape indexes based on facial curvature characteristics and perform dynamic programming. Experimental results show that the proposed method is capable of determining the angle of faces accurately over a wide range of poses. In addition, 96.8% face recognition rate has been achieved based on the proposed method with 300 individuals with seven different poses. Contact Information Hwanjong Song Email: [email protected] Contact Information Ukil Yang Email: [email protected] Contact Information Sangyoun Lee Email: [email protected] Contact Information Kwanghoon Sohn Email: [email protected] =================================== 16. King Hong Cheung, Adams Wai-Kin Kong, David Zhang, Mohamed Kamel, Jane Toby You: Revealing the Secret of FaceHashing. 106-112 Electronic Edition (link) BibTeX Revealing the Secret of FaceHashing King-Hong Cheung1 Contact Information, Adams Kong1, 2 Contact Information, David Zhang1 Contact Information, Mohamed Kamel2 Contact Information and Jane You1 Contact Information (1) Biometrics Research Centre, Department of Computing, The Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong (2) Pattern Analysis and Machine Intelligence Lab, University of Waterloo, 200 University Avenue West, Ontario, Canada Abstract Biometric authentication has attracted substantial attention over the past few years. It has been reported recently that a new technique called FaceHashing, which is proposed for personal authentication using file:///C|/Documents%20and%20Settings/Dorothea%20M....uments/My%20Webs/mysite11/BDG/BioBits2006-0103.html (23 of 83)1/6/2006 7:17:27 AM Biometric Bits 2006-01-No,03 face images, has achieved perfect accuracy and zero equal error rates (EER). In this paper, we are going to reveal that the secret of FaceHashing in achieving zero EER is based on a false assumption. This is done through simulating the claimants’ experiments. Thus, we would like to alert the use of “safe” token. Contact Information King-Hong Cheung Email: [email protected] Contact Information Adams Kong Email: [email protected] Contact Information David Zhang Email: [email protected] Contact Information Mohamed Kamel Email: [email protected] Contact Information Jane You Email: [email protected] =================================== 17. Manuele Bicego, Enrico Grosso, Massimo Tistarelli: Person Authentication from Video of Faces: A Behavioral and Physiological Approach Using Pseudo Hierarchical Hidden Markov Models. 113-120 Electronic Edition (link) BibTeX Person Authentication from Video of Faces: A Behavioral and Physiological Approach Using Pseudo Hierarchical Hidden Markov Models Manuele Bicego1, Enrico Grosso1 and Massimo Tistarelli2 (1) DEIR University of Sassari, via Torre Tonda 34 07100 Sassari, Italy (2) DAP University of Sassari, piazza Duomo 6 07041 Alghero (SS), Italy Abstract In this paper a novel approach to identity verification, based on the analysis of face video streams, is proposed, which makes use of both physiological and behavioral features. While physical features are obtained from the subject’s face appearance, behavioral features are obtained by asking the subject to vocalize a given sentence. The recorded video sequence is modelled using a Pseudo-Hierarchical Hidden Markov Model, a new type of HMM in which the emission probability of each state is represented by another HMM. The number of states are automatically determined from the data by unsupervised clustering of expressions of faces in the video. Preliminary results on real image data show the feasibility of the proposed approach. =================================== 18. Zongying Ou, Xusheng Tang, Tieming Su, Pengfei Zhao: Cascade AdaBoost Classifiers with Stage Optimization for Face Detection. 121-128 Electronic Edition (link) BibTeX Cascade AdaBoost Classifiers with Stage Optimization for Face Detection file:///C|/Documents%20and%20Settings/Dorothea%20M....uments/My%20Webs/mysite11/BDG/BioBits2006-0103.html (24 of 83)1/6/2006 7:17:27 AM Biometric Bits 2006-01-No,03 Zongying Ou1 Contact Information, Xusheng Tang1, Tieming Su1 and Pengfei Zhao1 (1) Key Laboratory for Precision and Non-traditional Machining Technology, of Ministry of Education, Dalian University of Technology, Dalian 116024, P.R. China Abstract In this paper, we propose a novel feature optimization method to build a cascade Adaboost face detector for real-time applications, such as teleconferencing, user interfaces, and security access control. AdaBoost algorithm selects a set of weak classifiers and combines them into a final strong classifier. However, conventional AdaBoost is a sequential forward search procedure using the greedy selection strategy, the weights of weak classifiers may not be optimized. To address this issue, we proposed a novel Genetic Algorithm post optimization procedure for a given boosted classifier, which yields better generalization performance. Contact Information Zongying Ou Email: [email protected] =================================== 19. Jooyoung Park, Daesung Kang, James T. Kwok, Sang-Woong Lee, Bon-Woo Hwang, Seong-Whan Lee: Facial Image Reconstruction by SVDD-Based Pattern De-noising. 129-135 Electronic Edition (link) BibTeX Facial Image Reconstruction by SVDD-Based Pattern De-noising Jooyoung Park1, Daesung Kang1, James T. Kwok2, Sang-Woong Lee3, Bon-Woo Hwang3 and SeongWhan Lee3 (1) Department of Control and Instrumentation Engineering, Korea University, Jochiwon, Chungnam, 339-700, Korea (2) Department of Computer Science, Hong Kong University of Science and Technology, Clear Water Bay, Hong Kong (3) Department of Computer Science and Engineering, Korea University, Anam-dong, Seongbuk-ku, Seoul 136-713, Korea Abstract The SVDD (support vector data description) is one of the most well-known one-class support vector learning methods, in which one tries the strategy of utilizing balls defined on the feature space in order to distinguish a set of normal data from all other possible abnormal objects. In this paper, we consider the problem of reconstructing facial images from the partially damaged ones, and propose to use the SVDD-based de-noising for the reconstruction. In the proposed method, we deal with the shape and texture information separately. We first solve the SVDD problem for the data belonging to the given prototype facial images, and model the data region for the normal faces as the ball resulting from the SVDD problem. Next, for each damaged input facial image, we project its feature vector onto the decision boundary of the SVDD ball so that it can be tailored enough to belong to the normal region. Finally, we obtain the image of the reconstructed face by obtaining the pre-image of the projection, and then further processing with its shape and texture information. The applicability of the proposed method is illustrated via some experiments dealing with damaged facial images. file:///C|/Documents%20and%20Settings/Dorothea%20M....uments/My%20Webs/mysite11/BDG/BioBits2006-0103.html (25 of 83)1/6/2006 7:17:27 AM Biometric Bits 2006-01-No,03 =================================== 20. Xiujuan Chai, Shiguang Shan, Laiyun Qing, Wen Gao: Pose Estimation Based on Gaussian Error Models. 136-143 Electronic Edition (link) BibTeX Pose Estimation Based on Gaussian Error Models Xiujuan Chai1 Contact Information, Shiguang Shan2 Contact Information, Laiyun Qing2 Contact Information and Wen Gao1, 2 Contact Information (1) School of Computer Science and Technology, Harbin Institute of Technology, 150001 Harbin, China (2) ICT-ISVISION Joint R&D Lab for Face Recognition, ICT, CAS, 100080 Beijing, China Abstract In this paper, a new method is presented to estimate the 3D pose of facial image based on statistical Gaussian error models. The basic idea is that the pose angle can be computed by the orthogonal projection computation if the specific 3D shape vector of the given person is known. In our algorithm, Gaussian probability density function is used to model the distributions of the 3D shape vector as well as the errors between the orthogonal projection computation and the weak perspective projection. By using the prior knowledge of the errors distribution, the most likely 3D shape vector can be referred by the labeled 2D landmarks in the given facial image according to the maximum posterior probability theory. Refining the error term, thus the pose parameters can be estimated by the transformed orthogonal projection formula. Experimental results on real images are presented to give the objective evaluation. =================================== 21. Zhong Jin, Franck Davoine, Zhen Lou, Jingyu Yang: A Novel PCA-Based Bayes Classifier and Face Analysis. 144-150 Electronic Edition (link) BibTeX A Novel PCA-Based Bayes Classifier and Face Analysis Zhong Jin1, 2 Contact Information, Franck Davoine3 Contact Information, Zhen Lou2 and Jingyu Yang2 Contact Information (1) Centre de Visió per Computador, Universitat Autònoma de Barcelona, Barcelona, Spain (2) Department of Computer Science, Nanjing University of Science and Technology, Nanjing, People’s Republic of China (3) HEUDIASYC CNRS Mixed Research Unit, Compiègne University of Technology, 60205 Compiègne cedex, France Abstract The classical Bayes classifier plays an important role in the field of pattern recognition. Usually, it is not easy to use a Bayes classifier for pattern recognition problems in high dimensional spaces. This paper proposes a novel PCA-based Bayes classifier for pattern recognition problems in high dimensional spaces. Experiments for face analysis have been performed on CMU facial expression image database. It is shown that the PCA-based Bayes classifier can perform much better than the minimum distance classifier. And, with the PCA-based Bayes classifier, we can obtain a better understanding of data. Contact Information Zhong Jin file:///C|/Documents%20and%20Settings/Dorothea%20M....uments/My%20Webs/mysite11/BDG/BioBits2006-0103.html (26 of 83)1/6/2006 7:17:27 AM Biometric Bits 2006-01-No,03 Email: [email protected] Contact Information Franck Davoine Email: [email protected] Contact Information Jingyu Yang Email: [email protected] =================================== 22. Stan Z. Li, Rufeng Chu, Meng Ao, Lun Zhang, Ran He: Highly Accurate and Fast Face Recognition Using Near Infrared Images. 151-158 Electronic Edition (link) BibTeX Highly Accurate and Fast Face Recognition Using Near Infrared Images Stan Z. Li1 Contact Information, RuFeng Chu1 Contact Information, Meng Ao1 Contact Information, Lun Zhang1 Contact Information and Ran He1 Contact Information (1) Center for Biometrics and Security Research & National Laboratory of Pattern Recognition, Institute of Automation, Chinese Academy of Sciences, 95 Zhongguancun Donglu Beijing 100080, China Abstract In this paper, we present a highly accurate, realtime face recognition system for cooperative user applications. The novelties are: (1) a novel design of camera hardware, and (2) a learning based procedure for effective face and eye detection and recognition with the resulting imagery. The hardware minimizes environmental lighting and delivers face images with frontal lighting. This avoids many problems in subsequent face processing to a great extent. The face detection and recognition algorithms are based on a local feature representation. Statistical learning is applied to learn most effective features and classifiers for building face detection and recognition engines. The novel imaging system and the detection and recognition engines are integrated into a powerful face recognition system. Evaluated in real-world user scenario, a condition that is harder than a technology evaluation such as Face Recognition Vendor Tests (FRVT), the system has demonstrated excellent accuracy, speed and usability. This work was supported by Chinese National 863 Projects 2004AA1Z2290 & 2004AA119050. See: http://research.microsoft.com/iccv2005/demo/StanLi/IR-Face-Demo.pdf Contact Information Stan Z. Li URL: http://www.cbsr.ia.ac.cn Contact Information RuFeng Chu URL: http://www.cbsr.ia.ac.cn Contact Information Meng Ao URL: http://www.cbsr.ia.ac.cn Contact Information Lun Zhang URL: http://www.cbsr.ia.ac.cn file:///C|/Documents%20and%20Settings/Dorothea%20M....uments/My%20Webs/mysite11/BDG/BioBits2006-0103.html (27 of 83)1/6/2006 7:17:27 AM Biometric Bits 2006-01-No,03 Contact Information Ran He URL: http://www.cbsr.ia.ac.cn =================================== 23. Jaewon Sung, Daijin Kim: Background Robust Face Tracking Using Active Contour Technique Combined Active Appearance Model. 159-165 Electronic Edition (link) BibTeX Background Robust Face Tracking Using Active Contour Technique Combined Active Appearance Model Jaewon Sung1 Contact Information and Daijin Kim1 Contact Information (1) Biometrics Engineering Research Center (BERC), Pohang University of Science and Technology, Abstract This paper proposes a two stage AAM fitting algorithm that is robust to the cluttered background and a large motion. The proposed AAM fitting algorithm consists of two alternative procedures: the active contour fitting to find the contour sample that best fits the face image and then the active appearance model fitting over the best selected contour. Experimental results show that the proposed active contour based AAM provides better accuracy and convergence characteristics in terms of RMS error and convergence rate, respectively, than the existing robust AAM. This work was supported by the Korea Science and Engineering Foundation (KOSEF) through the Biometrics Engineering Research Center (BERC) at Yonsei University. Contact Information Jaewon Sung Email: [email protected] Contact Information Daijin Kim Email: [email protected] =================================== 24. Hui Kong, Xuchun Li, Jian-Gang Wang, Chandra Kambhamettu: Ensemble LDA for Face Recognition. 166-172 Electronic Edition (link) BibTeX Ensemble LDA for Face Recognition Hui Kong1, Xuchun Li1, Jian-Gang Wang2 and Chandra Kambhamettu3 (1) School of Electrical and Electronic Engineering, Nanyang Technological University, 50 Nanyang Ave.,639798, Singapore (2) Institute for Infocomm Research, 21 Heng Mui Keng Terrace,119613, Singapore (3) Department of Computer and Information Science, University of Delaware,Newark, DE 197162712, Abstract Linear Discriminant Analysis (LDA) is a popular feature extraction technique for face image recognition file:///C|/Documents%20and%20Settings/Dorothea%20M....uments/My%20Webs/mysite11/BDG/BioBits2006-0103.html (28 of 83)1/6/2006 7:17:27 AM Biometric Bits 2006-01-No,03 and retrieval. However, It often suffers from the small sample size problem when dealing with the high dimensional face data. Two-step LDA (PCA+LDA) [1][2][3] is a class of conventional approaches to address this problem. But in many cases, these LDA classifiers are overfitted to the training set and discard some useful discriminative information. In this paper, by analyzing the overfitting problem for the two-step LDA approach, a framework of Ensemble Linear Discriminant Analysis (EnLDA) is proposed for face recognition with small number of training samples. In EnLDA, a Boosting-LDA (BLDA) and a Random Sub-feature LDA (RS-LDA) schemes are incorporated together to construct the total weak-LDA classifier ensemble. By combining these weak-LDA classifiers using majority voting method, recognition accuracy can be significantly improved. Extensive experiments on two public face databases verify the superiority of the proposed EnLDA over the state-of-the-art algorithms in recognition accuracy. =================================== 25. Enrique Argones-Rúa, Josef Kittler, José Luis Alba-Castro, Daniel González-Jiménez: Information Fusion for Local Gabor Features Based Frontal Face Verification. 173-181 Electronic Edition (link) BibTeX Information Fusion for Local Gabor Features Based Frontal Face Verification Enrique Argones Rúa1, Josef Kittler2, Jose Luis Alba Castro1 and Daniel González Jiménez1 (1) Signal Theory Group, Signal Theory and Communications Dep., University of Vigo, 36310, Spain (2) Centre for Vision, Speech and Signal Processing, University of Surrey, Guildford GU2 7XH, UK Abstract We address the problem of fusion in a facial component approach to face verification. In our study the facial components are local image windows defined on a regular grid covering the face image. Gabor jets computed in each window provide face representation. A fusion architecture is proposed to combine the face verification evidence conveyed by each facial component. A novel modification of the linear discriminant analysis method is presented that improves fusion performance as well as providing a basis for feature selection. The potential of the method is demonstrated in experiments on the XM2VTS data base. The references of this article are secured to subscribers. =================================== 26. Sreekar Krishna, John Black, Sethuraman Panchanathan: Using Genetic Algorithms to Find Person-Specific Gabor Feature Detectors for Face Indexing and Recognition. 182-191 Electronic Edition (link) BibTeX Using Genetic Algorithms to Find Person-Specific Gabor Feature Detectors for Face Indexing and Recognition Sreekar Krishna1 Contact Information, John Black1 and Sethuraman Panchanathan1 (1) Center for Cognitive Ubiquitous Computing (CUbiC), Arizona State University, Tempe AZ85281, file:///C|/Documents%20and%20Settings/Dorothea%20M....uments/My%20Webs/mysite11/BDG/BioBits2006-0103.html (29 of 83)1/6/2006 7:17:27 AM Biometric Bits 2006-01-No,03 Abstract In this paper, we propose a novel methodology for face recognition, using person-specific Gabor wavelet representations of the human face. For each person in a face database a genetic algorithm selects a set of Gabor features (each feature consisting of a particular Gabor wavelet and a corresponding (x, y) face location) that extract facial features that are unique to that person. This set of Gabor features can then be applied to any normalized face image, to determine the presence or absence of those characteristic facial features. Because a unique set of Gabor features is used for each person in the database, this method effectively employs multiple feature spaces to recognize faces, unlike other face recognition algorithms in which all of the face images are mapped into a single feature space. Face recognition is then accomplished by a sequence of face verification steps, in which the query face image is mapped into the feature space of each person in the database, and compared to the cluster of points in that space that represents that person. The space in which the query face image most closely matches the cluster is used to identify the query face image. To evaluate the performance of this method, it is compared to the most widely used subspace method for face recognition: Principle Component Analysis (PCA). For the set of 30 people used in this experiment, the face recognition rate of the proposed method is shown to be substantially higher than PCA.In this paper, we propose a novel methodology for face recognition, using person-specific Gabor wavelet representations of the human face. For each person in a face database a genetic algorithm selects a set of Gabor features (each feature consisting of a particular Gabor wavelet and a corresponding (x, y) face location) that extract facial features that are unique to that person. This set of Gabor features can then be applied to any normalized face image, to determine the presence or absence of those characteristic facial features. Because a unique set of Gabor features is used for each person in the database, this method effectively employs multiple feature spaces to recognize faces, unlike other face recognition algorithms in which all of the face images are mapped into a single feature space. Face recognition is then accomplished by a sequence of face verification steps, in which the query face image is mapped into the feature space of each person in the database, and compared to the cluster of points in that space that represents that person. The space in which the query face image most closely matches the cluster is used to identify the query face image. To evaluate the performance of this method, it is compared to the most widely used subspace method for face recognition: Principle Component Analysis (PCA). For the set of 30 people used in this experiment, the face recognition rate of the proposed method is shown to be substantially higher than PCA. Contact Information Sreekar Krishna Email: [email protected] =================================== 27. Bingpeng Ma, Fei Yang, Wen Gao, Baochang Zhang: The Application of Extended Geodesic Distance in Head Poses Estimation. 192-198 Electronic Edition (link) BibTeX The Application of Extended Geodesic Distance in Head Poses Estimation Bingpeng Ma1, 3, Fei Yang1, 3, Wen Gao1, 2, 3 and Baochang Zhang2 (1) Institute of Computing Technology, Chinese Academy of Sciences, Beijing 100080, China (2) Department of Computer Science and Engineering, Harbin Institute of Technology, Harbin, 150001, China (3) Graduate School of the Chinese Academy of Sciences, Beijing 100039, China Abstract This paper we proposes an extended geodesic distance for head pose estimation. In ISOMAP, two approaches are applied for neighborhood construction, called k-neighbor and ε-neighbor. For the kneighbor, the number of the neighbors is a const k. For the other one, all the distances between the neighbors is less than ε. Either the k-neighbor or the ε-neighbor neglects the difference of each point. This paper proposes an new method called the kc-neighbor, in which the neighbors are defined based on c time distance of the k nearest neighbor, which can avoid the neighborhood graph unconnected and improve the accuracy in computing neighbors. In this paper, SVM rather than MDS is applied to classify head poses after the geodesic distances are computed. The experiments show the effectiveness of the proposed method. =================================== 28. Bindang Xue, Wenfang Xue, Zhiguo Jiang: Improved Parameters Estimating Scheme for E-HMM with Application to Face Recognition. 199file:///C|/Documents%20and%20Settings/Dorothea%20M....uments/My%20Webs/mysite11/BDG/BioBits2006-0103.html (30 of 83)1/6/2006 7:17:27 AM Biometric Bits 2006-01-No,03 205 Electronic Edition (link) BibTeX Improved Parameters Estimating Scheme for E-HMM with Application to Face Recognition Bindang Xue1 Contact Information, Wenfang Xue2 Contact Information and Zhiguo Jiang1 Contact Information (1) Image processing center, Beihang University, Beijng 100083, China (2) Institute of Automation, Chinese Academy of Sciences, 100088, Beijing, China Abstract This paper presents a new scheme to initialize and re-estimate Embedded Hidden Markov Models(EHMM) parameters for face recognition. Firstly, the current samples were assumed to be a subset of the whole training samples, after the training process, the E-HMM parameters and the necessary temporary parameters in the parameter re-estimating process were saved for the possible retraining use. When new training samples were added to the training samples, the saved E-HMM parameters were chosen as the initial model parameter. Then the E-HMM was retrained based on the new samples and the new temporary parameters were obtained. Finally, these temporary parameters were combined with saved temporary parameters to form the final E-HMM parameters for representing one person face. Experiments on ORL databases show the improved method is effective. =================================== 29. Cuiping Zhang, Fernand S. Cohen: Component-Based Active Appearance Models for Face Modelling. 206-212 Electronic Edition (link) BibTeX Component-Based Active Appearance Models for Face Modelling Cuiping Zhang1 Contact Information and Fernand S. Cohen1 Contact Information (1) Eletrical and Computer Engineering Department, Drexel University, Philadelphia PA 19104, USA Abstract The Active Appearance Model (AAM) is a powerful tool for modelling a class of objects such as faces. However, it is common to see a far from optimal local alignment when attempting to model a face that is quite different from training faces. In this paper, we present a novel component-based AAM algorithm. By modelling three components inside the face area, then combining them with a global AAM, face alignment achieves both local as well as global optimality. We also utilize local projection models to locate face contour points. Compared to the original AAM, our experiment shows that this new algorithm is more accurate in shape localization as the decoupling allows more flexibility. Its insensitivity to different face background patterns is also clearly manifested. Contact Information Cuiping Zhang Email: [email protected] Contact Information Fernand S. Cohen Email: [email protected] =================================== file:///C|/Documents%20and%20Settings/Dorothea%20M....uments/My%20Webs/mysite11/BDG/BioBits2006-0103.html (31 of 83)1/6/2006 7:17:27 AM Biometric Bits 2006-01-No,03