Mihaela Gordan

Learn More
Visual speech recognition is an emerging research field. In this paper, we examine the suitability of support vector machines for visual speech recognition. Each word is modeled as a temporal sequence of visemes corresponding to the different phones realized. One support vector machine is trained to recognize each viseme and its output is converted to a(More)
In this paper we proposed a visual speech recognition network based on Support Vector Machines. Each word of the dictionary is modeled by a set of temporal sequences of visemes. Each viseme is described by a support vector machine, and the temporal character of speech is modeled by integrating the support vector machines as nodes into Viterbi decoding(More)
Our objective is to investigate image processing directly in the compressed domain, without full decompression. Compressed domain image processing algorithms provide a powerful computational alternative to classical (pixel level) based implementations. The field is just emerging and the algorithms reported in the literature are mostly based on linear(More)
In this paper we propose a visual speech recognition network based on Support Vector Machines. Each word of the dictionary is described as a temporal sequence of visemes. Each viseme is described by a support vector machine, and the temporal character of speech is modeled by integrating the support vector machines as nodes into a Viterbi decoding lattice.(More)
Many existing works in face recognition are based solely on visible images. The use of bimodal systems based on visible and thermal images is seldom reported in face recognition, despite its advantage of combining the discriminative power of both modalities, under expressions or pose variations. In this paper, we investigate the combined advantages of(More)
The problem of lip contour detection is critical in the lipreading systems based on contour processing. The typical contour detection strategy based on image segmentation in homogeneous regions fails in the case when the mouth images available for lipreading are low-contrast gray level images. Most of the solutions adopted require manual marking of some(More)
Medical volume segmentation in various imaging modalities using real 3D approaches (in contrast to slice-by-slice segmentation) represents an actual trend. The increase in the acquisition resolution leads to large amount of data, requiring solutions to reduce the dimensionality of the segmentation problem. In this context, the real-time interaction with the(More)
—The perspective projection models the way a 3D scene is transformed into a 2D image, usually through a camera or an eye. In a projective transformation, parallel lines intersect in a point called vanishing point. This paper presents in detail two calibration methods that exploit the properties of vanishing points. The aim of the paper is to offer a(More)