No description available.

Dans cette thèse, nous développons une approche supervisée de reconnaissance d’objets basée sur l’utilisation de nouveaux descripteurs d’images globaux inspirés du modèle du cortex visuel humain primaire V1 en tant que groupe de roto-translations semi-discrètes SE (2,N)=R² x ZN produit semi-direct entre R² et ZN. La méthode proposée est basée sur des descripteurs de Fourier généralisés et rotationnels définis sur le groupe SE (2,N), qui sont invariants aux transformations géométriques (translations, et rotations). De plus, nous montrons que ces descripteur de Fourier sont faiblement complets, dans le sens qu’ils permettent de discriminer sur un ensemble ouvert et dense L² (SE(2,N)) de fonctions à support compact, donc distinguer entre des images réelles. Ces descripteurs sont ensuite utilisés pour alimenter un classifieur de type SVM dans le cadre de la reconnaissance d’objets. Nous avons mené une séries d’expérimentations dans le but d’évaluer notre méthode sur les bases de visages RL, CVL et ORL et sur la base d’images d’objets variés COIL-100, et de comparer ses performances à celles des méthodes basées sur des descripteurs globaux et locaux. Les résultats obtenus ont montré que notre approche est en mesure de concurrencer de nombreuses techniques de reconnaissance d’objets existantes et de surpasser de nombreuse autres. Ces résultats ont également montré que notre méthode est robuste aux bruits. Enfin, nous avons employé la technique proposée pour reconnaître des navires dans un contexte de surveillance maritime. / In this thesis, we develop a supervised object recognition method using new global image descriptors inspired by the model of the human primary visual cortex V1. Mathematically speaking, the latter is modeled as the semi-discrete roto-translation group SE (2,N)=R² x ZN semi-direct product between R² and ZN. Therefore, our technique is based on generalized and rotational Fourier descriptors defined in SE (2,N) , and which are invariant to natural geometric transformations (translations, and rotations). Furthermore, we show that such Fourier descriptors are weakly complete, in the sense that they allow to distinguish over an open and dense set of compactly supported functions in L² (SE(2,N)) , hence between real-world images. These descriptors are later used in order to feed a Support Vector Machine (SVM) classifier for object recognition purposes. We have conducted a series of experiments aiming both at evaluating and comparing the performances of our method against existing both local - and global - descriptor based state of the art techniques, using the RL, the CVL, and the ORL face databases, and the COIL-100 image database (containing various types of objects). The obtained results have demonstrated that our approach was able to compete with many existing state of the art object recognition techniques, and to outperform many others. These results have also shown that our method is robust to noise. Finally, we have applied the proposed method on vessels recognition in the framework of maritime surveillance.

Stimuli outside classical receptive fields have been shown to exert significant influence over the activities of neurons in primary visual cortexWe propose that contextual influences are used for pre-attentive visual segmentation, in a new framework called segmentation without classification. This means that segmentation of an image into regions occurs without classification of features within a region or comparison of features between regions. This segmentation framework is simpler than previous computational approaches, making it implementable by V1 mechanisms, though higher leve l visual mechanisms are needed to refine its output. However, it easily handles a class of segmentation problems that are tricky in conventional methods. The cortex computes global region boundaries by detecting the breakdown of homogeneity or translation invariance in the input, using local intra-cortical interactions mediated by the horizontal connections. The difference between contextual influences near and far from region boundaries makes neural activities near region boundaries higher than elsewhere, making boundaries more salient for perceptual pop-out. This proposal is implemented in a biologically based model of V1, and demonstrated using examples of texture segmentation and figure-ground segregation. The model performs segmentation in exactly the same neural circuit that solves the dual problem of the enhancement of contours, as is suggested by experimental observations. Its behavior is compared with psychophysical and physiological data on segmentation, contour enhancement, and contextual influences. We discuss the implications of segmentation without classification and the predictions of our V1 model, and relate it to other phenomena such as asymmetry in visual search.

Diss. (sammanfattning) Linköping : Linköpings universitet, 2009. / Härtill 5 uppsatser.

No description available.

Affect and cognition have traditionally been considered mutually exclusive domains and their study has evolved into two separate research fields. In recent years, however, there is increasing evidence of affective modulations of cognitive processes and interest in the study of affective cognition has grown. This thesis presents analyses of data collected in four mixed-design experiments between 2009 and 2011, which were designed to investigate affective memory and its electrophysiological correlates, individual differences in said affective memory and electrophysiological correlates, the time-course of affective memory and attentional disengagement from affective stimuli respectively. The first aim of the research presented here was to further understanding of how affective content influences picture processing and memory. Event-related potentials (ERPs) provide a valuable tool for the investigation of modulations of cognitive processes, as their excellent temporal resolution allows for the dissociation between different processes contributing to behavioural outcomes. Several important results for the study of affective cognition are reported: The late positive potential (LPP) was shown to be modulated differentially by affective content when compared to a behavioural attentional disengagement task. While the behavioural measure of attention replicated findings from participants’ self-report of arousal, LPP enhancement did not. This novel finding demonstrates that the affective modulation of the LPP cannot be used as an electrophysiological marker of slowed attentional disengagement as is common in the literature. In the domain of recognition memory, affective modulation of performance was shown to be time-sensitive, with effects developing faster for negative than for positive picture content. Affective pictures were associated with a less conservative response bias than neutral pictures but only negative pictures elicited better discrimination performance, driven by an increased in the rate of “remembered” as compared to merely familiar pictures. This was reflected in an increase of the ERP old/new effect for negative pictures in the 500 to 800ms time window, the purported correlate of recollection. The late right-frontal old/new effect between 800 and 1500 ms post stimulus onset was shown to be attenuated by affective content, supporting the interpretation of the late right-frontal effect as a correlate of relevance detection over a retrieval success interpretation. In combination, the findings add weight to the conclusion that affective content enhances memory through selective memory sparing for affective stimuli. Novel evidence for gender differences in affective cognition was found. Comparisons between female and male participants revealed that the affective modulation of the late right-frontal effect differs between the genders, underlining the importance of assessing and understanding gender differences as part of the study of affective cognition. Brain-derived neurotrophic factor (BDNF) gene val66met single nucleotide polymorphism (SNP), a small genetic change that affects the functioning of BDNF, a protein that plays an important role in neuron growth, differentiation and survival, is shown here to also affect the interaction of affect and cognition. BDNF val66met genotype modulated the early “familiarity” old/new effect selectively in response to positive pictures. The present study clearly demonstrates the value of the ERP technique in the investigation of individual differences in affective and cognitive processing and the need to take such individual differences into account as part of the endeavour to fully understand the mechanisms of affective processing, cognition and affective cognition. A better understanding of the role of gender and genetic differences in the affective modulation of affective processing and memory will have important practical implications in fields where affect and cognition interact.

Orientadores: Roberto José Maria Covolan, Rickson Coelho Mesquita / Tese (doutorado) - Universidade Estadual de Campinas, Instituto de Física Gleb Wataghin / Made available in DSpace on 2018-08-25T21:27:17Z (GMT). No. of bitstreams: 1 Anjos_CarlosAlessandroSilvados_D.pdf: 14228439 bytes, checksum: 3f8ec7a93c1a21bbd6147d80351b5c9b (MD5) Previous issue date: 2014 / Resumo: A atividade cerebral envolve uma complexa rede de processos neurofisiológicos cuja demanda energética requer constante aporte de glicose e oxigênio, supridos através da microcirculação sanguínea cerebral. Variações na circulação sanguínea cerebral decorrentes de ativação neuronal podem ser detectadas e registradas através da técnica óptica denominada NIRS (near infrared spectroscopy). Neste trabalho, desenvolvemos estudos sobre a aplicabilidade da técnica NIRS sob três diferentes abordagens, envolvendo indivíduos saudáveis e portadores de uma condição patológica específica. Os estudos com indivíduos saudáveis foram focados na aplicabilidade da técnica NIRS para investigar alterações hemodinâmicas associadas à estimulação do córtex visual e à realização de exercício físico. No caso envolvendo condição patológica, investigamos a aplicabilidade clínica desta mesma técnica para avaliar a perfusão cerebral de pacientes com estenose carotídea. Nos estudos sobre estimulação do córtex visual, buscou-se estabelecer uma relação entre a frequência do estímulo apresentado ao voluntário e parâmetros obtidos das curvas hemodinâmicas. Nos experimentos associados a exercícios físicos, ciclistas semiprofissionais e indivíduos fisicamente ativos realizaram testes em bicicletas simulando tipos específicos de provas, envolvendo tarefas abertas (teste progressivo) e fechadas (teste contra relógio em uma prova de 4Km), em diferentes condições (controle, ingestão de placebo ou cafeína), buscando caracterizar diferenças em aspectos da hemodinâmica cerebral a elas associadas. Por fim, nos experimentos envolvendo indivíduos com estenose carotídea, foram realizadas medidas de NIRS, durante a realização de testes de apneia com duração de 30 segundos, buscando estabelecer a aplicabilidade da técnica na avaliação clínica deste tipo de patologia. Para cada uma dessas abordagens, foram estabelecidos parâmetros associados às respostas hemodinâmicas obtidas através de tarefas e estímulos específicos, que permitiram caracterizar e quantificar os processos fisiológicos envolvidos em cada tipo de experimento, demonstrando assim a aplicabilidade da técnica NIRS para o estudo da atividade cerebral sob as condições experimentais em questão / Abstract: Brain activity involves a complex network of neurophysiological processes whose energy demand requires constant supply of glucose and oxygen which is provided by the cerebral microcirculation. Changes in cerebral blood flow due to neuronal activation can be detected and recorded by the optical technique called NIRS (near infrared spectroscopy). In this work, we developed studies on the applicability of the NIRS technique under three different approaches, involving healthy subjects and patients with a specific pathological condition. Studies in healthy subjects were focused on the applicability of the NIRS technique to investigate hemodynamic changes associated with stimulation of the visual cortex and the performance of physical exercise. In the case involving pathological condition, we investigated the clinical applicability of this same technique to evaluate cerebral perfusion in patients with carotid stenosis. In studies of stimulation of the visual cortex, we sought to establish a relationship between the frequency of the stimulus presented to volunteer and parameters obtained from the hemodynamic curves. In the experiments associated with exercise, semi-professional cyclists and physically active subjects performed tasks on bicycles simulating specific types of tests, involving open tasks (progressive test) and closed (a 4km test against the clock), under different conditions (control, placebo or caffeine intake), seeking to characterize differences in aspects of cerebral hemodynamics associated with them. Finally, in experiments involving individuals with carotid stenosis, NIRS measurements were carried out during apnea tests lasting 30 seconds, from which we sought to establish the applicability of the technique in the clinical evaluation of this type of pathology. For each of these approaches, parameters associated with the hemodynamic responses obtained by stimulation of specific tasks allowed to characterize and quantify the physiological processes involved in each type of experiment, thus demonstrating the applicability of the NIRS technique to the study of brain activity under the experimental conditions in question / Doutorado / Física / Doutor em Ciências

Dissimilarity plays a very important role in object recognition. But, finding perceptual dissimilarity between objects is non-trivial as it is not equivalent to the pixel dissimilarity between the objects (For example, two white noise images appear very similar even when they have different intensity values at every corresponding pixel). However, visual search allows us to reliably measure perceptual dissimilarity between a pair of objects. When the target object is dissimilar to the distracter, visual search becomes easy and it will be difficult otherwise. Even though we can measure perceptual dissimilarity between objects, we still do not understand either the underlying mechanisms or the visual features involved in the computation of dissimilarities. For this thesis, I have explored perceptual dissimilarity in two studies – by looking at known simple features and understanding how they combine, and using computational models to understand or discover complex features. In the first study, we looked at how dissimilarity between two simple objects with known features can be predicted using dissimilarities between individual features. Specifically, we investigated how search for targets differing in multiple features (intensity, length, orientation) from the distracters is related to searches for targets differing in each of the individual features. We found that multiple feature dissimilarities could be predicted as a linear combination of individual feature dissimilarities. Also, we demonstrated for the first time that Aspect ratio of the object emerges as a novel feature in visual search. This work has been published in the Journal of Vision (Pramod & Arun, 2014). Having established in the first study that simple features combine linearly, we devised a second study to investigate dissimilarities in complex shapes. Since it is known that shape is one of the salient and complex features in object representation, we chose silhouettes of animals and abstract objects to explore the nature of dissimilarity computations. We conducted visual search using pairs of these silhouettes on humans to get an estimate of perceptual dissimilarity. We then used various computational models of shape representation (like Fourier Descriptors, Curvature Scale Space, HMAX model etc) to see how well they can predict the observed dissimilarities. We found that many of these computational models were able to predict the perceptual dissimilarities of a large number of object pairs. However, we also observed many cases where computational models failed to predict perceptual dissimilarities. The manuscript related to this study is under preparation.

L'intégration sensorielle nécessite un alignement correct des cartes nerveuses dans le cerveau. Dans les couches superficielles du colliculus supérieur, situé dans le mésencéphale, des projections en provenance de la rétine ainsi que du cortex visuel primaire doivent être alignés, mais les mécanismes sous-jacents de ce processus demeurent à ce jour méconnus. Afin d'élucider ces mécanismes, éphrine-A3 a été sur-exprimée dans un modèle murin, dans une sous-population des cellules ganglionnaires de la rétine, induisant un disruption de l'alignement de la carte rétino-colliculaire sur la carte cortico-colliculaire. L'inactivation in vivo d'éphrine-A3 dans la rétine restaure un phénotype sauvage. Une analyse théorique utilisant un modèle informatique a permis la modélisation des donnés obtenues. Ces résultats ont permis l'identification d'un principe de base dans l'alignement des cartes et des mécanismes associés, validés par un modèle théorique. / Efficient sensory processing requires correct alignment of neural maps throughout the brain. In the superficial layers of the superior colliculus in the midbrain, projections from retinal ganglion cells and V1 cortex must be aligned to form a visuotopic map, but the basic principle and underlying mechanism are elusive and still incomplete. In a new mouse model, over-expression of ephrin-A3 in a subset of retinal ganglion cells disrupts the cortico-collicular map alignment onto the retino-collicular map, creating a visuotopic mismatch. In vivo inactivation of retinal ephrin-A3 over-expression restores a wild-type corticocollicular map. Theoretical analyses using an original algorithm models the stochastic nature of maps formation and alignment, and recapitulates our observations. Our results identify a basic principle for the alignment of converging maps and the associated mechanism, validated by a theoretical model.

Face recognition that is invariant to pose and illumination is a problem solved effortlessly by the human brain, but the computational details that underlie such efficient recognition are still far from clear. This thesis draws on research from psychology and neuroscience about face and object recognition and the visual system in order to develop a novel computational method for face detection, feature selection and representation, and memory structure for recall. A biologically plausible framework for developing a face recognition system will be presented. This framework can be divided into four parts: 1) A face detection system. This is an improved version of a biologically inspired feedforward neural network that has modifiable connections and reflects the hierarchical and elastic structure of the visual system. The face detection system can detect if a face is present in an input image, and determine the region which contains that face. The system is also capable of detecting the pose of the face. 2) A face region selection mechanism. This mechanism is used to determine the Gabor-style features corresponding to the detected face, i.e., the features from the region of interest. This region of interest is selected using a feedback mechanism that connects the higher level layer of the feedforward neural network where ultimately the face is detected to an intermediate level where the Gabor style features are detected. 3) A face recognition system which is based on the binary encoding of the Gabor style features selected to represent a face. Two alternative coding schemes are presented, using 2 and 4 bits to represent a winning orientation at each location. The effectiveness of the Gabor-style features and the different coding schemes in discriminating faces from different classes is evaluated using the Yale B Face Database. The results from this evaluation show that this representation is close to other results on the same database. 4) A theoretical approach for a memory system capable of memorising sequences of poses. A basic network for memorisation and recall of sequences of labels have been implemented, and from this it is extrapolated a memory model that could use the ability of this model to memorise and recall sequences, to assist in the recognition of faces by memorising sequences of poses. Finally, the capabilities of the detection and recognition parts of the system are demonstrated using a demo application that can learn and recognise faces from a webcam.