Results 21 - 30
of
34
Texture and Shape Information Fusion for Facial Action Unit Recognition
- in First International Conference on Advances in Computer-Human Interaction (ACHI
, 2008
"... Abstract—A novel method that fuses texture and shape infor-mation to achieve Facial Action Unit (FAU) recognition from video sequences is proposed. In order to extract the texture information, a subspace method based on Discriminant Non-negative Matrix Factorization (DNMF) is applied on the dif-fere ..."
Abstract
-
Cited by 1 (0 self)
- Add to MetaCart
(Show Context)
Abstract—A novel method that fuses texture and shape infor-mation to achieve Facial Action Unit (FAU) recognition from video sequences is proposed. In order to extract the texture information, a subspace method based on Discriminant Non-negative Matrix Factorization (DNMF) is applied on the dif-ference images of the video sequence, calculated taking under consideration the neutral and the most expressive frame, to extract the desired classification label. The shape information consists of the deformed Candide facial grid (more specifically the grid node displacements between the neutral and the most expressive facial expression frame) that corresponds to the facial expression depicted in the video sequence. The shape information is afterwards classified using a two-class Support Vector Machine (SVM) system. The fusion of texture and shape information is performed using Median Radial Basis Functions (MRBFs) Neural Networks (NNs) in order to detect the set of present FAUs. The accuracy achieved in the Cohn-Kanade database is equal to 92.1 % when recognizing the 17 FAUs that are responsible for facial expression development.
TITLE: A New Incremental Classification Approach: Monitoring The Risk of Heart Disease
"... ii ..."
(Show Context)
of the Requirements
"... instilled much stronger personality in me than I could ever have hoped for. I could not have achieved much at UTA were it not for the support given to me by certain individuals, who I will like to acknowledge. My foremost appreciation goes to Dr. Michael Manry, my advisor, who has been a great sourc ..."
Abstract
- Add to MetaCart
(Show Context)
instilled much stronger personality in me than I could ever have hoped for. I could not have achieved much at UTA were it not for the support given to me by certain individuals, who I will like to acknowledge. My foremost appreciation goes to Dr. Michael Manry, my advisor, who has been a great source of inspiration throughout my graduate school years. He saw the work from inception to fruition and provided all the help to make this work possible. I admire his subject expertise, contribution and devotion to the field of Neural Networks, and incessant help to his students in various forms viz. teaching, regular laboratory visits, immediate feedback and motivation to better understand the field of Neural Networks. He patiently directed me through my research, and taught me all I know about critical reasoning and analysis, presentation of work done and even basic requirements. I believe he is the best counselor one can ever hope for. I thank Dr Stephen Gibbs and Dr. Bernard Svihel for reviewing my work and also for agreeing to serve on my thesis committee. The EE coursework at UTA and my undergraduate institute Gujarat University in India provided the fillip and toolkit to weave the pieces of this work together. Hence a sincere thanks to all the teachers who selflessly strive to spread education to whom I dedicate this work. ii Finally, I must express my sincere gratitude to my family for all their love and support. My parents, Mr. Vallabh Shah and Mrs. Smita Shah, for all their love, faith and support. My sister and my brother-in-law, Jeny and Jashmin, for being there for me whenever I needed them. Specially Mr. Jashmin Shah, whose ideas influenced my way of thinking and changed my approach towards graduate studies and research activities. He has been and will be a role model for me.
RADAR
"... Abstract—This paper develops a maximum a posteriori (MAP) probability estimation framework for shape-from-shading (SFS) from synthetic aperture radar (SAR) images. The aim is to use this method to reconstruct surface topography from a single radar image of relatively complex terrain. Our MAP framewo ..."
Abstract
- Add to MetaCart
(Show Context)
Abstract—This paper develops a maximum a posteriori (MAP) probability estimation framework for shape-from-shading (SFS) from synthetic aperture radar (SAR) images. The aim is to use this method to reconstruct surface topography from a single radar image of relatively complex terrain. Our MAP framework makes explicit how the recovery of local surface orientation depends on the whereabouts of terrain edge features and the available radar reflectance information. To apply the resulting process to real world radar data, we require probabilistic models for the appearance of terrain features and the relationship between the orientation of surface normals and the radar reflectance. We show that the SAR data can be modeled using a Rayleigh-Bessel distribution and use this distribution to develop a maximum likelihood algorithm for detecting and labeling terrain edge features. Moreover, we show how robust statistics can be used to estimate the characteristic parameters of this distribution. We also develop an empirical model for the SAR reflectance function. Using the reflectance model, we perform Lambertian correction so that a conventional SFS algorithm can be applied to the radar data. The initial surface normal direction is constrained to point in the direction of the nearest ridge or ravine feature. Each surface normal must fall within a conical envelope whose axis is in the direction of the radar illuminant. The extent of the envelope depends on the corrected radar reflectance and the variance of the radar signal statistics. We explore various ways of smoothing the field of surface normals using robust statistics. Finally, we show how to reconstruct the terrain surface from the smoothed field of surface normal vectors. The proposed algorithm is applied to various SAR data sets containing relatively complex terrain structure. Index Terms—Synthetic aperture radar imagining, shape-from-shading, terrain surface reconstruction, maximum a posteriori probability estimation, robust statistics. æ
Clasificación de Cáncer Cérvico Uterino mediante la Red
"... RESUMEN.- En este trabajo, la red neuronal Función de Base Radial de Rango Tipo M (RMRBF) es usada para la clasificación de imágenes digitales celulares de cáncer cérvico uterino. Los resultados experimentales obtenidos indican que la red propuesta es mejor en comparación con la red RBF en términos ..."
Abstract
- Add to MetaCart
(Show Context)
RESUMEN.- En este trabajo, la red neuronal Función de Base Radial de Rango Tipo M (RMRBF) es usada para la clasificación de imágenes digitales celulares de cáncer cérvico uterino. Los resultados experimentales obtenidos indican que la red propuesta es mejor en comparación con la red RBF en términos de capacidad de clasificación.
unknown title
"... for the classification of Pap smear microscopic images. Simulation results indicate that the proposed neural network consistently outperforms the RBF network in terms of classification capabilities. KEYWORDS.- Neural networks, Papanicolaou test, Pap Smear, RMRBF, Cervical Cancer. I. ..."
Abstract
- Add to MetaCart
(Show Context)
for the classification of Pap smear microscopic images. Simulation results indicate that the proposed neural network consistently outperforms the RBF network in terms of classification capabilities. KEYWORDS.- Neural networks, Papanicolaou test, Pap Smear, RMRBF, Cervical Cancer. I.
REFERENCES
"... of the four problems with a perfect classification record for all bit strings of finite lengths. Induction is seen here as the process of deriving a stable metric space to separate the training groups. A stable metric space is one containing well-separated, compact clusters. From the perspectives of ..."
Abstract
- Add to MetaCart
(Show Context)
of the four problems with a perfect classification record for all bit strings of finite lengths. Induction is seen here as the process of deriving a stable metric space to separate the training groups. A stable metric space is one containing well-separated, compact clusters. From the perspectives of clustering and statistical discriminant analysis, the proposed theory provides the most meaningful and methodological clustering/separation criterion because it goes beyond the limitations of the Euclidean space to the metric space, and beyond the limitations of the fixed space to a dynamic selection from an infinite family of spaces. The proposed model is based upon a modified version of the model in [8]. The modifications are deceptively subtle but the consequences are profound because now an elegant, computational (discrete), analytical (continuous) and systematic method is being offered for the difficult unsupervised pattern learning problem.
Variational Learning for Gaussian Mixture Models
"... Abstract—This paper proposes a joint maximum likelihood and Bayesian methodology for estimating Gaussian mixture models. In Bayesian inference, the distributions of parameters are modeled, characterized by hyperparameters. In the case of Gaussian mixtures, the distributions of parameters are conside ..."
Abstract
- Add to MetaCart
(Show Context)
Abstract—This paper proposes a joint maximum likelihood and Bayesian methodology for estimating Gaussian mixture models. In Bayesian inference, the distributions of parameters are modeled, characterized by hyperparameters. In the case of Gaussian mixtures, the distributions of parameters are considered as Gaussian for the mean, Wishart for the covariance, and Dirichlet for the mixing probability. The learning task consists of estimating the hyperparameters characterizing these distributions. The integration in the parameter space is decoupled using an unsupervised variational methodology entitled variational expectation–maximization (VEM). This paper introduces a hyperparameter initialization procedure for the training algorithm. In the first stage, distributions of parameters resulting from successive runs of the expectation–maximization algorithm are formed. Afterward, maximum-likelihood estimators are applied to find appropriate initial values for the hyperparameters. The proposed initialization provides faster convergence, more accurate hyperparameter estimates, and better generalization for the VEM training algorithm. The proposed methodology is applied in blind signal detection and in color image segmentation. Index Terms—Bayesian inference, expectation–maximization algorithm, Gaussian mixtures, maximum log-likelihood estimation,
Multilayer Perceptron as the basis for Gaming Motion Generation
"... The computer gaming industry is large and growing rapidly. Consumers demand realistic motion in computer games. Consequently, the ability to realistically simulate motion is vital in computer games development. Currently, most simulation of motion is done through the use of a physics engine which es ..."
Abstract
- Add to MetaCart
(Show Context)
The computer gaming industry is large and growing rapidly. Consumers demand realistic motion in computer games. Consequently, the ability to realistically simulate motion is vital in computer games development. Currently, most simulation of motion is done through the use of a physics engine which essentially involves numerically solution of the set of differential equations which describe the corresponding physics situation in the real world. This paper proposes an alternative strategy using a multilayer perceptron to generate these simulations. Benchmarking against traditional physics engines shows that there are considerable advantages to the new methodology. 1 Background Traditionally, methods such as key framing, motion capture, and high-level control have been used for the generation of motion in computer games [26]. More recently, a technique known as Physics Based Modelling (PBM) has been developed [11]. PBM is widely considered to be superior to previous methods [30]. This is because PBM allows easy generation of families of similar motions, is able to describe realistic, complex, real world animation such as metal fracture, cloth drapping, explosions and rigid body deformation, and can generate reproducible motions [21] and [27]. It has this power because it expresses motion by using Newton's Laws, thus more accurately represents the real world [15]. In effect, the objects within a PBM system follow the same physical rules that we do. Physics is applied to the generation of motion by using what are known as `physics engines' [21]. Physics engines are used as a layer in 2D/3D game software that applies physical laws to generate the realistic motion of objects in the virtual world [14] and [23]. Figure 1 illustrates the way that physics engines are interfaced wit...
detection and removal of cracks
"... Digital image processing techniques for the ..."
(Show Context)
