Les recherches menées au cours de cette thèse de Doctorat s'inscrivent dans les activités du laboratoire commun OPERA (OPtique EmbaRquée Active) impliquant ESSILOR-LUXOTTICA et le CNRS. L’objectif est de contribuer au développement des “lunettes du futur” intégrant des fonctions d'obscurcissement, de focalisation ou d'affichage qui s’adaptent en permanence à la scène et au regard de l’utilisateur. Ces nouveaux dispositifs devront être dotés de capacités de perception, de décision et d’action, et devront respecter des contraintes d'encombrement, de poids, de consommation énergétique et de temps de traitement. Ils présentent par conséquent des connexions évidentes avec la robotique. Dans ce contexte, les recherches ont consisté à investiguer la structure et la construction de tels systèmes afin d’identifier leurs enjeux et difficultés. Pour ce faire, la première tâche a été de mettre en place des émulateurs de divers types de lunettes actives, qui permettent de prototyper et d’évaluer efficacement diverses fonctions. Dans cette phase de prototypage et de test, ces émulateurs s’appuient naturellement sur une architecture logicielle modulaire typique de la robotique. La seconde partie de la thèse s'est focalisée sur le prototypage d’un composant clé des lunettes du futur, qui implique une contrainte supplémentaire de basse consommation : le système de suivi du regard, aussi appelé oculomètre. Le principe d’un assemblage de photodiodes et d’un traitement par réseau de neurones a été proposé. Un simulateur a été mis au point, ainsi qu’une étude de l'influence de l'agencement des photodiodes et de l’hyper-paramétrisation du réseau sur les performances de l'oculomètre. / The research carried out during this doctoral thesis takes place within the OPERA joint laboratory (OPtique EmbaRquée Active) involving ESSILOR-LUXOTTICA and the CNRS. The aim is to contribute to the development of "glasses of the future", which feature obscuration, focus or display capabilities that continuously adapt to the scene and the user gaze. These new devices will be endowed with perception, decision and action capabilities, and will have to respect constraints of space, weight, energy consumption and processing time. They therefore show obvious connections with robotics. In this context, the structure and building of such systems has been investigated in order to identify their issues and difficulties. To that end, the first task was to set up emulators of various types of active glasses, which enable the prototyping and effective testing of various functions. In this prototyping and testing phase, these emulators naturally rely on a modular software architecture typical of robotics. The second part of the thesis focused on the prototyping of a key component which implies an additional constraint on low consumption, namely the eye tracking system, also known as gaze tracker. The principle of a photodiode assembly and of a neural network processing has been proposed. A simulator has been developed, as well as a study of the influence of the arrangement of photodiodes and the hyper-parametrization of the network on the performance of the oculometer.

Purpose Early detection of abnormal visual acuity (VA) is crucial in the identification and management of ocular and visual abnormalities in infants. Currently, the Teller Acuity Cards (TACs) are considered the gold standard for clinical testing and are effective in obtaining a quick estimate of an infant’s VA, but they have certain drawbacks. They rely on a subjective assessment of the baby’s looking behavior. Despite this, TACs have been found to have good validity and repeatability. The current study investigates a new method to objectively assess visual acuity in infants, which is uses a video gaze tracker (GT) and computer-generated stimuli, developed in the lab of M. Eizenman at the University of Toronto. The purpose was to validate this method in adults and infants against current clinical VA tests. Visual scanning patterns were measured by the GT system that requires minimal subject cooperation in adult and infant populations. The targets were judged as seen when the relative fixation time on the grating exceeded a pre-determined threshold, as compared to the fixation time on the luminance-matched background. Methods Experiment 1: In 15 uncorrected myopic adults, binocular grating VA was measured. The targets were square-wave gratings of spatial frequency ranging from 2.3 to 37 cpd presented randomly in one of four positions on the screen. There were 6 objective protocols (in which VA was judged by fixations). The subjects were naïve, as the only instruction given to the participants was to look towards the screen. The experimenter, who presented the gratings also acted as an observer by making judgments of seen/not seen responses using the objective information provided by the software. Objective GT VA was compared with VA measured with subjective responses using the same stimuli and with Teller Acuity Cards (TACs). Experiment 2: Binocular grating VA for horizontal gratings was measured in 20 typically-developing infants aged 3 to 12 months. Spatial frequency ranged from 0.32 to 42 cpd and VA was measured on two visits with both the GT and TACs. A staircase protocol was used to obtain the VA threshold in the GT. The experimenter controlled the staircase method and an observer used the objective information of visual fixations using the software to judge if the grating was seen or not. Video cartoons were shown between stimulus presentations to keep the infant’s attention towards the screen. VA was also measured with the TACs held in the vertical orientation, so that the gratings were horizontal, similar to the GT method. A TAC stage was specially designed with a vertical slot in which the cards could be presented. The observer was masked regarding the participant’s age and the starting spatial frequency. The study co-ordinator determined the choice of the start card which was randomized between participants so as to give an equal number of participants with each start card. The same start card was used for the second session of each infant. The threshold was defined as the highest spatial frequency for which the infant gave a clear, correct look and an unclear/inconsistent look for the next higher level. The observer, who was masked regarding the absolute spatial frequency, increased or decreased the spatial frequency until this threshold was determined. Results Experiment 1: The mean age of participants in the adult study was 28.47±7.93 yrs and their mean uncorrected logMAR acuity was 0.9±0.2. There was no obvious difference among the mean acuities obtained by 6 objective GT protocols, the subjective GT protocol and the TACs. The GT showed agreement of 93% and 100 % within half an octave compared with the subjective protocol and TACs (horizontal gratings) respectively. There was 100% agreement within 1 octave of the objective GT with both the subjective protocol and the TACs (horizontal gratings). The objective gaze tracker VA showed significant correlation with uncorrected refractive error (r =0.87, p < 0.001). Experiment 2: The mean age of participants was 7.9±2.5 months. In both visits, the testability of the TACs was 100% across all infants. GT had 100% testability on the first visit and 95% testability on the second visit. The mean TAC acuity over two visits for all the infants was 0.7±0.23 log cycles per degree, while the mean log GT acuity over two visits was 0.86±0.30. Infant GT VA acuity estimates were within 1 octave of the TACs 90% and 79% of the time for the first and second visit respectively, while GT VA estimates were within half octave of the TACs 63.2% and 47% of the time for the first and second visit respectively. Eighty-seven percent of the GT VAs and 72.5% of TACs were within one octave of the mean age norms, although on average the GT gave better acuities than the TACs. There was an increase in GT VA with increasing age (r=0.80, p<0.005 for the first visit and r=0.77, p<0.005 for the second visit). Both the TACs and the GT had repeatability of 89.5% within 1 octave between visits and 84.2% and 79% within half octave between visits respectively. Conclusions In adults, the gaze tracker gave VA thresholds which were equivalent to the TACs and were not significantly different from subjectively determined grating VA. The agreement of the GT with TACs in infants and with norms in the infant literature established good validity for the GT. Finally, the significant correlation with age confirmed the validity of the measurements of the gaze tracker. The repeatability of the gaze tracker was similar to that of the TACs, demonstrating the quality usefulness of the test. These results demonstrate the potential for an automated test of infant visual acuity, which could be a powerful and useful tool for visual acuity assessment in infants and other population groups who cannot respond verbally. The staircase protocol established in the study could be fully automated in an objective version of the test. The raw data of eye movements obtained in this study such as the pattern of first fixations, time taken for first fixations, time spent fixing the stimulus, typical stimulus duration and time between presentations could be used to develop algorithms for fully automated testing of VA in infants.

Purpose Early detection of abnormal visual acuity (VA) is crucial in the identification and management of ocular and visual abnormalities in infants. Currently, the Teller Acuity Cards (TACs) are considered the gold standard for clinical testing and are effective in obtaining a quick estimate of an infant’s VA, but they have certain drawbacks. They rely on a subjective assessment of the baby’s looking behavior. Despite this, TACs have been found to have good validity and repeatability. The current study investigates a new method to objectively assess visual acuity in infants, which is uses a video gaze tracker (GT) and computer-generated stimuli, developed in the lab of M. Eizenman at the University of Toronto. The purpose was to validate this method in adults and infants against current clinical VA tests. Visual scanning patterns were measured by the GT system that requires minimal subject cooperation in adult and infant populations. The targets were judged as seen when the relative fixation time on the grating exceeded a pre-determined threshold, as compared to the fixation time on the luminance-matched background. Methods Experiment 1: In 15 uncorrected myopic adults, binocular grating VA was measured. The targets were square-wave gratings of spatial frequency ranging from 2.3 to 37 cpd presented randomly in one of four positions on the screen. There were 6 objective protocols (in which VA was judged by fixations). The subjects were naïve, as the only instruction given to the participants was to look towards the screen. The experimenter, who presented the gratings also acted as an observer by making judgments of seen/not seen responses using the objective information provided by the software. Objective GT VA was compared with VA measured with subjective responses using the same stimuli and with Teller Acuity Cards (TACs). Experiment 2: Binocular grating VA for horizontal gratings was measured in 20 typically-developing infants aged 3 to 12 months. Spatial frequency ranged from 0.32 to 42 cpd and VA was measured on two visits with both the GT and TACs. A staircase protocol was used to obtain the VA threshold in the GT. The experimenter controlled the staircase method and an observer used the objective information of visual fixations using the software to judge if the grating was seen or not. Video cartoons were shown between stimulus presentations to keep the infant’s attention towards the screen. VA was also measured with the TACs held in the vertical orientation, so that the gratings were horizontal, similar to the GT method. A TAC stage was specially designed with a vertical slot in which the cards could be presented. The observer was masked regarding the participant’s age and the starting spatial frequency. The study co-ordinator determined the choice of the start card which was randomized between participants so as to give an equal number of participants with each start card. The same start card was used for the second session of each infant. The threshold was defined as the highest spatial frequency for which the infant gave a clear, correct look and an unclear/inconsistent look for the next higher level. The observer, who was masked regarding the absolute spatial frequency, increased or decreased the spatial frequency until this threshold was determined. Results Experiment 1: The mean age of participants in the adult study was 28.47±7.93 yrs and their mean uncorrected logMAR acuity was 0.9±0.2. There was no obvious difference among the mean acuities obtained by 6 objective GT protocols, the subjective GT protocol and the TACs. The GT showed agreement of 93% and 100 % within half an octave compared with the subjective protocol and TACs (horizontal gratings) respectively. There was 100% agreement within 1 octave of the objective GT with both the subjective protocol and the TACs (horizontal gratings). The objective gaze tracker VA showed significant correlation with uncorrected refractive error (r =0.87, p < 0.001). Experiment 2: The mean age of participants was 7.9±2.5 months. In both visits, the testability of the TACs was 100% across all infants. GT had 100% testability on the first visit and 95% testability on the second visit. The mean TAC acuity over two visits for all the infants was 0.7±0.23 log cycles per degree, while the mean log GT acuity over two visits was 0.86±0.30. Infant GT VA acuity estimates were within 1 octave of the TACs 90% and 79% of the time for the first and second visit respectively, while GT VA estimates were within half octave of the TACs 63.2% and 47% of the time for the first and second visit respectively. Eighty-seven percent of the GT VAs and 72.5% of TACs were within one octave of the mean age norms, although on average the GT gave better acuities than the TACs. There was an increase in GT VA with increasing age (r=0.80, p<0.005 for the first visit and r=0.77, p<0.005 for the second visit). Both the TACs and the GT had repeatability of 89.5% within 1 octave between visits and 84.2% and 79% within half octave between visits respectively. Conclusions In adults, the gaze tracker gave VA thresholds which were equivalent to the TACs and were not significantly different from subjectively determined grating VA. The agreement of the GT with TACs in infants and with norms in the infant literature established good validity for the GT. Finally, the significant correlation with age confirmed the validity of the measurements of the gaze tracker. The repeatability of the gaze tracker was similar to that of the TACs, demonstrating the quality usefulness of the test. These results demonstrate the potential for an automated test of infant visual acuity, which could be a powerful and useful tool for visual acuity assessment in infants and other population groups who cannot respond verbally. The staircase protocol established in the study could be fully automated in an objective version of the test. The raw data of eye movements obtained in this study such as the pattern of first fixations, time taken for first fixations, time spent fixing the stimulus, typical stimulus duration and time between presentations could be used to develop algorithms for fully automated testing of VA in infants.

Augmented Reality (AR) devices are set apart from other mobile devices by the immersive experience they offer. Specifically, head-mounted AR devices can accurately sense and understand their environment through an increasingly powerful array of sensors such as cameras, depth sensors, eye gaze trackers, microphones, and inertial sensors. The ability of these devices to collect this information presents both challenges and opportunities to improve existing security and privacy techniques in this domain. Specifically, eye gaze tracking is a ready-made capability to analyze user intent, emotions, and vulnerability, and as an input mechanism. However, modern AR devices lack systems to address their unique security and privacy issues. Problems such as lacking local pairing mechanisms usable while immersed in AR environments, bystander privacy protections, and the increased vulnerability to shoulder surfing while wearing AR devices all lack viable solutions. In this dissertation, I explore how readily available eye gaze sensor data can be used to improve existing methods for assuring information security and protecting the privacy of those near the device. My research has presented three new systems, BystandAR, ShouldAR, and GazePair that each leverage user eye gaze to improve security and privacy expectations in or with Augmented Reality. As these devices grow in power and number, such solutions are necessary to prevent perception failures that hindered earlier devices. The work in this dissertation is presented in the hope that these solutions can improve and expedite the adoption of these powerful and useful devices. / Doctor of Philosophy / Augmented Reality (AR) devices are set apart from other mobile devices by the immersive experience they offer. The ability of these devices to collect information presents challenges and opportunities to improve existing security and privacy techniques in this domain. In this dissertation, I explore how readily available eye gaze sensor data can be used to improve existing methods for assuring security and protecting the privacy of those near the device. My research has presented three new systems, BystandAR, ShouldAR, and GazePair that each leverage user eye gaze to improve security and privacy expectations in or with Augmented Reality. As these devices grow in power and number, such solutions are necessary to prevent perception failures that hindered earlier devices. The work in this dissertation is presented in the hope that these solutions can improve and expedite the adoption of these powerful and useful devices.

Gaze tracking technology is widely used in Human Computer Interaction applications such as in interfaces for assisting people with disabilities and for driver attention monitoring. However, commercially available gaze trackers are expensive and their performance deteriorates if the user is not positioned in front of the camera and facing it. Also, head motion or being far from the device degrades their accuracy. This thesis focuses on the development of real-time time appearance based gaze tracking algorithms using low cost devices, such as a webcam or Kinect. The proposed algorithms are developed by considering accuracy, robustness to head pose variation and the ability to generalise to different persons. In order to deal with head pose variation, we propose to estimate the head pose and then compensate for the appearance change and the bias to a gaze estimator that it introduces. Head pose is estimated by a novel method that utilizes tensor-based regressors at the leaf nodes of a random forest. For a baseline gaze estimator we use an SVM-based appearance-based regressor. For compensating the appearance variation introduced by the head pose, we use a geometric model, and for compensating for the bias we use a regression function that has been trained on a training set. Our methods are evaluated on publicly available datasets.

<p>Gaze tracking is the estimation of the point in space a person is “looking at”. This is widely used in both diagnostic and interactive applications, such as visual attention studies and human-computer interaction. The most common commercial solution used to track gaze today uses a combination of infrared illumination and one or more cameras. These commercial solutions are reliable and accurate, but often expensive. The aim of this thesis is to construct a simple single-camera gaze tracker from off-the-shelf components. The method used for gaze tracking is based on infrared illumination and a schematic model of the human eye. Based on images of reflections of specific light sources in the surfaces of the eye the user’s gaze point will be estimated. Evaluation is also performed on both the software and hardware components separately, and on the system as a whole. Accuracy is measured in spatial and angular deviation and the result is an average accuracy of approximately one degree on synthetic data and 0.24 to 1.5 degrees on real images at a range of 600 mm.</p>

This thesis describes a research study in the field of Human-Computer Interaction (HCI), specifically usability evaluation. The research investigated ways to optimise the usability of Web sites. It specifically compared Flash and HTML versions of several different types of Web sites. The study commenced with a literature review regarding the process of usability evaluation of Web sites. Various usability evaluation methods and techniques were explored, and two emerging techniques were chosen for further investigation: (1) a contingent heuristic approach; and (2) eye gaze tracking. In order to confirm that these two techniques can be used effectively for Web site usability evaluations, two experiments were conducted to evaluate the usability of Web sites. The first experiment utilised an online questionnaire derived from the Website Usability Contingent Evaluation Tool (WUCET), which was based on the contingent heuristic approach. The second experiment involved eye gaze tracking with the faceLAB system, while participants interacted with Web sites of different types. Both experiments utilised Flash and HTML versions of the same set of Web sites. By analysing data collected from the experiments, comparisons between the usability of Flash and HTML versions were made. The results from quantitative and qualitative analyses of survey responses suggested that Flash version of Web sites, in general, provided higher usability than HTML version of Web sites, but eye gaze tracking data analyses showed no significant difference between the two versions. However, analyses of the eye tracking data were useful for improving understanding of the ways in which users interact with different versions of the Web sites. In addition, other influential factors that could affect the perceived usability of the Web sites, such as user’s gender and previous experience with computers and the Web, were also considered. The results of the experiments showed that in regard to Flash and HTML implementations of Web sites, there was a difference in Web site usability perception patterns between male and female users, and also between users with long-term computer/Web experience and users with short-term experience. In addition, a range of different types (purposes) of Web sites were utilised. In this study, selected Web sites fall into three broad categories according to their main purpose: (1) information; (2) entertainment; and (3) e-commerce. It was discovered that the type of Web sites also influenced the usability of Flash and HTML versions of Web sites, as perceived by users.

No description available.

<p>Gaze tracking means to detect and follow the direction in which a person looks. This can be used in for instance human-computer interaction. Most existing systems illuminate the eye with IR-light, possibly damaging the eye. The motivation of this thesis is to develop a truly non-intrusive gaze tracking system, using only a digital camera, e.g. a web camera.</p><p>The approach is to detect and track different facial features, using varying image analysis techniques. These features will serve as inputs to a neural net, which will be trained with a set of predetermined gaze tracking series. The output is coordinates on the screen.</p><p>The evaluation is done with a measure of accuracy and the result is an average angular deviation of two to four degrees, depending on the quality of the image sequence. To get better and more robust results, a higher image quality from the digital camera is needed.</p>

In this thesis, a visual 3-D eye gaze tracker is designed and implemented to tested via computer simulations and on an experimental setup. Proposed tracker is designed to examine human perception on autostereoscopic displays when the viewer is 3m away from such displays. Two different methods are proposed for calibrating personal parameters and gaze estimation, namely line of gaze (LoG) and line of sight (LoS) solutions. 2-D and 3-D estimation performances of the proposed system are observed both using computer simulations and the experimental setup. In terms of 2-D and 3-D performance criteria, LoS solution generates slightly better results compared to that of LoG on experimental setup and their performances are found to be comparable in simulations. 2-D estimation inaccuracy of the system is obtained as smaller than 0.5&deg / during simulations and approximately 1&deg / for the experimental setup. 3-D estimation inaccuracy of the system along x- and y-axis is obtained as smaller than 2&deg / during the simulations and the experiments. However, estimation accuracy along z-direction is significantly sensitive to pupil detection and head pose estimation errors. For typical error levels, 20cm inaccuracy along z-direction is observed during simulations, whereas this inaccuracy reaches 80cm in the experimental setup.