Jang Kim-fung. / Thesis (M.Phil.)--Chinese University of Hong Kong, 2003. / Includes bibliographical references (leaves 108-124). / Abstracts in English and Chinese. / Abstract --- p.i / Acknowledgement --- p.iii / Chapter 1 --- Introduction --- p.1 / Chapter 1.1 --- Background --- p.1 / Chapter 1.2 --- Introduction about Face Processing Framework --- p.4 / Chapter 1.2.1 --- Basic architecture --- p.4 / Chapter 1.2.2 --- Face detection --- p.5 / Chapter 1.2.3 --- Face tracking --- p.6 / Chapter 1.2.4 --- Face recognition --- p.6 / Chapter 1.3 --- The scope and contributions of the thesis --- p.7 / Chapter 1.4 --- The outline of the thesis --- p.8 / Chapter 2 --- Facial Feature Representation --- p.10 / Chapter 2.1 --- Facial feature analysis --- p.10 / Chapter 2.1.1 --- Pixel information --- p.11 / Chapter 2.1.2 --- Geometry information --- p.13 / Chapter 2.2 --- Extracting and coding of facial feature --- p.14 / Chapter 2.2.1 --- Face recognition --- p.15 / Chapter 2.2.2 --- Facial expression classification --- p.38 / Chapter 2.2.3 --- Other related work --- p.44 / Chapter 2.3 --- Discussion about facial feature --- p.48 / Chapter 2.3.1 --- Performance evaluation for face recognition --- p.49 / Chapter 2.3.2 --- Evolution of the face recognition --- p.52 / Chapter 2.3.3 --- Evaluation of two state-of-the-art face recog- nition methods --- p.53 / Chapter 2.4 --- Problem for current situation --- p.58 / Chapter 3 --- Face Detection Algorithms and Committee Ma- chine --- p.61 / Chapter 3.1 --- Introduction about face detection --- p.62 / Chapter 3.2 --- Face Detection Committee Machine --- p.64 / Chapter 3.2.1 --- Review of three approaches for committee machine --- p.65 / Chapter 3.2.2 --- The approach of FDCM --- p.68 / Chapter 3.3 --- Evaluation --- p.70 / Chapter 4 --- Facial Feature Localization --- p.73 / Chapter 4.1 --- Algorithm for gray-scale image: template match- ing and separability filter --- p.73 / Chapter 4.1.1 --- Position of face and eye region --- p.74 / Chapter 4.1.2 --- Position of irises --- p.75 / Chapter 4.1.3 --- Position of lip --- p.79 / Chapter 4.2 --- Algorithm for color image: eyemap and separa- bility filter --- p.81 / Chapter 4.2.1 --- Position of eye candidates --- p.81 / Chapter 4.2.2 --- Position of mouth candidates --- p.83 / Chapter 4.2.3 --- Selection of face candidates by cost function --- p.84 / Chapter 4.3 --- Evaluation --- p.85 / Chapter 4.3.1 --- Algorithm for gray-scale image --- p.86 / Chapter 4.3.2 --- Algorithm for color image --- p.88 / Chapter 5 --- Face Processing System --- p.92 / Chapter 5.1 --- System architecture and limitations --- p.92 / Chapter 5.2 --- Pre-processing module --- p.93 / Chapter 5.2.1 --- Ellipse color model --- p.94 / Chapter 5.3 --- Face detection module --- p.96 / Chapter 5.3.1 --- Choosing the classifier --- p.96 / Chapter 5.3.2 --- Verifying the candidate region --- p.97 / Chapter 5.4 --- Face tracking module --- p.99 / Chapter 5.4.1 --- Condensation algorithm --- p.99 / Chapter 5.4.2 --- Tracking the region using Hue color model --- p.101 / Chapter 5.5 --- Face recognition module --- p.102 / Chapter 5.5.1 --- Normalization --- p.102 / Chapter 5.5.2 --- Recognition --- p.103 / Chapter 5.6 --- Applications --- p.104 / Chapter 6 --- Conclusion --- p.106 / Bibliography --- p.107

O objetivo deste estudo foi determinar as variações de volume pós-operatórias associadas à cirurgia ortognática, com base nas variáveis tempo, gênero, idade e lado da face. Foram avaliadas fotografias tridimensionais de dezoito pacientes submetidos simultaneamente às osteotomias Le Fort I e sagital do ramo mandibular bilateral. As imagens foram obtidas por meio do aparelho Vectra® M3 (Canfield, NJ, EUA) e capturadas nos períodos pós-operatórios de: uma semana (1S), um mês (1M), dois meses (2M), seis meses (6M) e um ano (1A). A foto obtida no período 1A foi utilizada como base de comparação para as demais (1S, 1M, 2M e 6M). O maior volume facial foi observado 1 semana após a realização da cirurgia ortognática (47,11 cm3, DP = 30,5 cm3), e neste período os homens apresentaram valores mais elevados (67,27 cm3, DP = 33,54 cm3) em relação as mulheres (30,98 cm3, DP = 15,25 cm3) e o lado direito (26,58 cm3, DP = 17,22 cm3) maior que o esquerdo (20,52 cm3, DP = 14,93 cm3). A regressão do volume mais expressiva ocorreu entre a primeira semana e o primeiro mês de pós-operatório (58,9%). Não foram observadas diferenças significantes entre as faixas etárias analisadas. / The purpose of this retrospective study was to determine postoperative volume differences associated with orthognathic surgery, controlling for time, gender, age and side of face. Three-dimensional (3D) photographs were analyzed of eighteen patients submitted to simultaneous Le Fort I + Bilateral Sagittal Split osteotomies. The images were obtained using a Vectra® M3 device (Canfield, NJ, USA) and were captured postoperatively, at one week (1S), one month (1M), two months (2M), six months (6M) and one year (1A). The photo taken at 1A was used as the basis of comparison for the other photos (1S, 1M, 2M, 6M). Greatest facial volume was observed at one week after surgery (47,11 cm3, DP = 30,5 cm3), with men in this period showing higher values (67,27 cm3, DP = 33,54 cm3) than women (30,98 cm3, DP = 15,25 cm3) and the right side (26,58 cm3, DP = 17,22 cm3) higher than the left (20,52 cm3, DP = 14,93 cm3). The most extensive facial volume resolution occurred between the first week and the first month; postoperatively (58,9%). No significant difference was found between age groups.

Neste estudo, analisamos ocorrências de Dissensão em discussões entre falantes alemães com referência à sua expressão lingüística e aos procedimentos utilizados para manter o bom andamento de uma interação, os quais denominamos Trabalho da Face. Um elenco de Estratégias de Trabalho da Face foi definido com base nas quatro super-estratégias propostas por BROWN & LEVINSON, a saber: expressão direta sem compensação, compensação baseada na polidez positiva, compensação baseada na polidez negativa e expressão off record. A Dissensão foi definida como uma atitude proposicional de desacordo por parte do falante, referente a conteúdos ou atitudes imputados por ele ao seu interlocutor. A partir de seu escopo, definimos quatro categorias de Dissensão, sendo que tais categorias provaram-se influentes quanto à seleção das Estratégias de Trabalho da Face utilizadas para sua expressão. Em geral, a super-estratégia menos utilizada é a expressão direta sem compensação, seguida do uso de estratégias off record, observando-se ainda um notável equilíbrio no uso de estratégias de Compensação Positiva e Negativa. Quanto à expressão lingüística da Dissensão, concluiu-se que não há meios lingüísticos exclusivos para esse fim, porém, a grande maioria das ocorrências no corpus apresenta elementos lexicais semanticamente negativos. / This study analyses tokens of Dissension in samples of discussions among German speakers regarding its linguistic expression and the procedures used to maintain the due course of an interaction, which GOFFMAN denominated face-work. A set of face-work strategies was defined based upon the four super strategies of face-work proposed by BROWN & LEVINSON: on record, positive redress, negative redress and off record. Dissension was defined as a propositional attitude of a speaker referring to topics or attitudes imputed by him to his partner in conversation. According to its scope, Dissension was classified in four categories, which proved themselves influent upon the choice of the face-work strategies used to express them. In general, the least represented strategy is the Dissension bold on record, followed by strategies off record. A remarkable balance between positive and negative redress must be noted as well. As to the linguistic expression of Dissension, it has been concluded that there are no specific linguistic means to that, although the great majority of the occurrences in the corpus displays semantically negative lexical elements.

O objetivo deste estudo foi comparar, por meio de tomografia computadorizada de feixe cônico, a inclinação axial dos dentes posteriores e a morfologia óssea alveolar mandibular entre indivíduos com diferentes padrões faciais. Para isso, foram utilizadas 58 tomografias de indivíduos maiores de 14 anos de idade não tratados ortodonticamente. Os indivíduos foram divididos em 3 grupos de acordo com o padrão facial avaliado pelo índice VERT de Ricketts: Grupo 1: composto por 18 indivíduos com padrão braquifacial, com idade média de 21,58 anos (d.p.=7,43); Grupo 2: composto por 23 indivíduos com padrão mesofacial, com idade média de 19,14 anos (d.p.=5,19); e Grupo 3: composto por 17 indivíduos com padrão dolicofacial, com idade média de 19,09 anos (d.p.=6,89). Foi realizada uma avaliação quantitativa das tábuas ósseas vestibular e lingual dos dentes pósteroinferiores, em tomografias de feixe cônico. As medidas utilizadas foram: altura do osso mandibular vestibular e lingual, largura mandibular cervical e média, inclinação mandibular, inclinação da face vestibular, largura do molar, angulação do molar e ângulo dente/osso. A comparação intergrupos foi realizada com a utilização do teste ANOVA a um critério de seleção e teste de Tukey quando necessário. Para verificar se existe correlação entre o padrão facial e as variáveis estudadas foi utilizado o teste de correlação de Pearson. Os resultados demonstraram que os indivíduos braquifaciais apresentaram uma menor inclinação mandibular quando comparados aos indivíduos mesofaciais e dolicofaciais. Houve diferença significante da altura mandibular vestibular entre os três grupos avaliados, sendo que esta foi menor no grupo braquifacial, intermediária no grupo mesofacial e maior no grupo dolicofacial. A altura mandibular lingual foi menor no grupo braquifacial em comparação aos grupos mesofacial e dolicofacial. A largura mandibular média foi maior nos indivíduos braquifaciais em relação aos mesofaciais. / The aim of this study was to compare, with cone beam computed tomography, the axial inclination of posterior teeth and mandibular alveolar bone morphology among individuals with different facial patterns. For this, 58 CT scans of untreated individuals orthodontically aged more than 14 years were used. The subjects were divided into 3 groups according to the facial pattern evaluated by the VERT index by Ricketts: Group 1 comprised 18 individuals with brachyfacial pattern, with a mean age of 21.58 years (s.d.=7.43); Group 2 comprised 23 individuals with mesofacial pattern, with a mean age of 19.14 years (s.d.=5.19); and Group 3 comprised 17 individuals with dolichofacial pattern, with a mean age of 19.09 years (s.d.=6.89). A quantitative assessment of the buccal and lingual bone plates of the mandibular posterior teeth were made in cone beam CT scans. The variables measured were: buccal and lingual mandibular height, cervical and middle mandibular width, mandibular inclination, inclination of the buccal surface, molar width, molar angle and tooth/bone angle. The intergroup comparison was performed by one-way ANOVA and Tukey test when necessary. To check whether there is a correlation between facial pattern and the variables studied, the Pearson correlation test was used. The results showed that the brachyfacial subjects had a smaller mandibular angle when compared to mesofacial and dolichofacial individuals. There was significant difference in the buccal mandibular height between the three groups, and this was smaller in the brachyfacial group, intermediate in the mesofacial group and larger in the dolichofacial group. Lingual mandibular height was smaller in brachyfacial group when compared to mesofacial and dolichofacial groups. The mean mandibular width was greater in brachyfacial than in mesofacial subjects.

A presente pesquisa tem como objetivo caracterizar os principais elementos que promovem a interação entre professor e alunos em sala de aula, por meio da análise da transcrição de gravações de aulas universitárias, realizadas em uma universidade pública do interior do Estado de São Paulo. Ao analisarmos o momento de interação comunicativa ocorrido em sala de aula, notamos uma constante preocupação de os professores preservarem a face perante os alunos e, ao mesmo tempo, promoverem a interação em sala de aula, para que os alunos interajam, sanem quaisquer dúvidas que possam surgir e sintam-se em um ambiente agradável ao aprendizado. Para realizarmos tal análise, partimos de conceitos abordados pela Análise da Conversação como turno, face, poder, interação, marcadores discursivos e formatos de aulas expositivas. Baseados nesses conceitos, analisamos as aulas gravadas e identificamos a ocorrência de elementos que promovem a interação entre professores e alunos, assim como os efeitos que os mesmos surtem durante o processo interacional. O corpus foi gravado e transcrito de acordo com as normas publicadas pelo Projeto da Norma Urbana Culta NURC-SP. / This research has the purpose of characterizing the main elements that promote the interaction between teachers and students in classrooms, through an analyses of universities classes records, occurred in a privet university of a city in the middle of São Paulo state. By analyzing communicative interactions moments in classrooms, we have noticed a constant worry from the teacher is on preserving their faces for the students, and, at the same time, on promoting an interaction in classroom, for the students interact, ask for doubts and feel in a pleasant local for learning. To make this analysis, we took concepts by Conversation Analysis as turn, face, power, interaction, discursive markers and classes types. According to this theory, we analyzed the recorded classes and identified elements that promote interaction between teachers and students and the effects of theses elements in the conversation process. The corpus was recorded and transcribed according to the rules published by Urban Educated Norm Project NURC SP.

Cheung, Kin Wang. / Thesis (M.Phil.)--Chinese University of Hong Kong, 2009. / Includes bibliographical references (leaves 76-82). / Abstract also in Chinese. / Chapter 1. --- INTRODUCTION --- p.1 / Chapter 1.1. --- Background --- p.1 / Chapter 1.1.1. --- Introduction to Biometrics --- p.1 / Chapter 1.1.2. --- Face Recognition in General --- p.2 / Chapter 1.1.3. --- A Typical Face Recognition System Architecture --- p.4 / Chapter 1.1.4. --- Face Recognition in Surveillance Cameras --- p.6 / Chapter 1.1.5. --- Face recognition under Pose Variation --- p.9 / Chapter 1.2. --- Motivation and Objectives --- p.11 / Chapter 1.3. --- Related Works --- p.13 / Chapter 1.3.1. --- Overview of Pose-invariant Face Recognition --- p.13 / Chapter 1.3.2. --- Standard Face Recognition Setting --- p.14 / Chapter 1.3.3. --- Multi-Probe Setting --- p.19 / Chapter 1.3.4. --- Multi-Gallery Setting --- p.21 / Chapter 1.3.5. --- Non-frontal Face Databases --- p.23 / Chapter 1.3.6. --- Evaluation Metrics --- p.26 / Chapter 1.3.7. --- Summary of Non-frontal Face Recognition Settings --- p.27 / Chapter 1.4. --- Proposed Methods for Non-frontal Face Recognition --- p.28 / Chapter 1.5. --- Thesis Organization --- p.30 / Chapter 2. --- PARTIAL ELASTIC BUNCH GRAPH MATCHING --- p.31 / Chapter 2.1. --- Introduction --- p.31 / Chapter 2.2. --- EBGM for Non-frontal Face Recognition --- p.31 / Chapter 2.2.1. --- Overview of Baseline EBGM Algorithm --- p.31 / Chapter 2.2.2. --- Modified EBGM for Non-frontal Face Matching --- p.33 / Chapter 2.3. --- Experiments --- p.35 / Chapter 2.3.1. --- Experimental Setup --- p.35 / Chapter 2.3.2. --- Experimental Results --- p.37 / Chapter 2.4. --- Discussions --- p.40 / Chapter 3. --- FACE RECOGNITION BY FRONTAL VIEW SYNTHESIS WITH CALIBRATED STEREO CAMERAS --- p.43 / Chapter 3.1. --- Introduction --- p.43 / Chapter 3.2. --- Proposed Method --- p.44 / Chapter 3.2.1. --- Image Rectification --- p.45 / Chapter 3.2.2. --- Face Detection --- p.49 / Chapter 3.2.3. --- Head Pose Estimation --- p.51 / Chapter 3.2.4. --- Virtual View Generation --- p.52 / Chapter 3.2.5. --- Feature Localization --- p.54 / Chapter 3.2.6. --- Face Morphing --- p.56 / Chapter 3.3. --- Experiments --- p.58 / Chapter 3.3.1. --- Data Collection --- p.58 / Chapter 3.3.2. --- Synthesized Results --- p.59 / Chapter 3.3.3. --- Experiment Setup --- p.60 / Chapter 3.3.4. --- Experiment Results on FERET database --- p.61 / Chapter 3.3.5. --- Experiment Results on CAS-PEAL-R1 database --- p.62 / Chapter 3.4. --- Discussions --- p.64 / Chapter 3.5. --- Summary --- p.66 / Chapter 4. --- "EXPERIMENTS, RESULTS AND OBSERVATIONS" --- p.67 / Chapter 4.1. --- Experiment Setup --- p.67 / Chapter 4.2. --- Experiment Results --- p.69 / Chapter 4.3. --- Discussions --- p.70 / Chapter 5. --- CONCLUSIONS --- p.74 / Chapter 6. --- BIBLIOGRAPHY --- p.76

Purpose Traditionally, researchers have relied on eliciting preferences through face-to-face interviews. Recently, there has been a shift towards using internet-based methods. Different methods of data collection may be a source of variation in the results. In this study, we compare the preferences for the Adult Social Care Outcomes Toolkit (ASCOT) service user measure elicited using best-worst scaling (BWS) via a face-to-face interview and an online survey. Methods Data were collected from a representative sample of the general population in England. The respondents (face-toface: n = 500; online: n = 1001) completed a survey, which included the BWS experiment involving the ASCOT measure. Each respondent received eight best-worst scenarios and made four choices (best, second best, worst, second worst) in each scenario. Multinomial logit regressions were undertaken to analyse the data taking into account differences in the characteristics of the two samples and the repeated nature of the data. Results We initially found a number of small significant differences in preferences between the two methods across all ASCOT domains. These differences were substantially reduced-from 15 to 5 out of 30 coefficients being different at the 5% Level-and remained small in value after controlling for differences in observable and unobservable characteristics of the two samples. Conclusions This comparison demonstrates that face-to-face and internet surveys may lead to fairly similar preferences for social care-related quality of life when differences in sample characteristics are controlled for. With or without a constant sampling frame, studies should carefully design the BWS exercise and provide similar levels of clarification to participants in each survey to minimise the amount of error variance in the choice process.

[Truncated abstract] Recent studies have provided evidence that face-coding mechanisms reference a norm or average face (Leopold, O`Toole, Vetter & Blanz, 2001; Rhodes & Jeffery, 2006). The central aim of this thesis was to establish whether distinct norms, and dissociable neural mechanisms code faces of different race and sex categories. Chapter 1 provides a brief introduction to norm based coding of faces, and reviews evidence for the existence of distinct norms for different races and sexes. Chapter 1 then introduces adaptation as a tool for investigating these ideas. Chapter 2 presents two adaptation studies that examined how faces of different races are coded. The aim of these studies was to determine whether dissociable neural mechanisms (or distinct face norms) code faces of different races. Chinese and Caucasian participants rated the normality of Caucasian and Chinese test faces, before and after adaptation to distorted faces of one race (e.g., 'contracted' Chinese faces; Experiment 1) or distorted faces of both races (e.g., 'contracted' Chinese faces and 'expanded' Caucasian faces; Experiment 2). Following adaptation to faces of one race, there were changes in perceived normality for faces of both races (i.e., perceptual aftereffects), indicating that common neural mechanisms code Chinese and Caucasian faces. However, aftereffects were significantly smaller in faces of the unadapted race suggesting some sensitivity to the race of faces. This sensitivity was also evident in Experiment 2. ... Some dissociability was also found in the coding of faces of different iv sexes. In Experiments 2 and 3, participants adapted to oppositely distorted faces of both sexes. Weak sex-selective aftereffects were found. Taken together, the findings suggest that male and female faces are coded by dissociable but not completely distinct neural populations. Chapter 4 examined whether the aftereffects reported for faces of different races or sexes reflected the adaptation of high-level neural mechanisms tuned to the social category information in faces, or earlier coding mechanisms tuned to simple physical differences between face groups. Chinese and Caucasian participants adapted to oppositely distorted face sets that were the same distance apart on a morph continua. The face sets were either from different race categories (e.g., contracted Chinese faces and expanded Caucasian faces), or from the same race category, (e.g., contracted Chinese faces and expanded caricatured Chinese faces). Larger opposite aftereffects were found when face sets were from different race categories, than when they were from the same race category suggesting that oppositely adapted neural mechanisms are tuned to social category differences rather than simple physical differences in faces. Together, these studies shed new light on how we code faces from different face categories. Specifically, the findings indicate that faces of different races and sexes are coded by both common and race- or sex-selective neural mechanisms. In addition, the findings are consistent with the possibility that race- and sex-selective norms and dimensions are used to code faces in face space. The implications of these findings and possible avenues for future research are discussed.

Schizophrenia is a psychiatric disorder characterized by social dysfunction. People with schizophrenia misinterpret social information and it is suggested that this difficulty may result from visual processing deficits. As faces are one of the most important sources of social information it is hypothesized that people suffering from the disorder have impairments in the visual face processing system. It is unclear which mechanism of the face processing system is impaired but two types of deficits are most often proposed: a deficit in face perception in general (i.e., processing of facial features as such) and a deficit in facial emotion processing (i.e., recognition of emotional facial expressions). Due to the contradictory evidence from behavioural, electrophysiological as well as neuroimaging studies offering support for the involvement of one or the other deficit in schizophrenia it is early to make any conclusive statements as to the nature and level of impairment. Further studies are needed for a better understanding of the key mechanism and abnormalities underlying social dysfunction in schizophrenia.

<p>Schizophrenia is a psychiatric disorder characterized by social dysfunction. People with schizophrenia misinterpret social information and it is suggested that this difficulty may result from visual processing deficits. As faces are one of the most important sources of social information it is hypothesized that people suffering from the disorder have impairments in the visual face processing system. It is unclear which mechanism of the face processing system is impaired but two types of deficits are most often proposed: a deficit in face perception in general (i.e., processing of facial features as such) and a deficit in facial emotion processing (i.e., recognition of emotional facial expressions). Due to the contradictory evidence from behavioural, electrophysiological as well as neuroimaging studies offering support for the involvement of one or the other deficit in schizophrenia it is early to make any conclusive statements as to the nature and level of impairment. Further studies are needed for a better understanding of the key mechanism and abnormalities underlying social dysfunction in schizophrenia.</p>