A. Blake and M. Isard. 3D position, attitude and shape input using video tracking of hands and lips. In Proc. Siggraph, pages 185--192, 1994.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....system has evolved to be so good at perceiving Johansson s stimuli. Perception of biological motion may be divided into two phases: rst detection and, possibly, segmentation; then tracking. Of the two, tracking has recently been object of much attention and considerable progress has been made [19, 18, 2, 8, 9, 3, 23, 7]. Detection (given two frames: is there a human, where ) on the contrary, remains an open problem. In this paper we address the problem of de ning and estimating a perceptual model of biological motion and use it for detecting the human body and labeling it in monocular image sequences. By ....

A. Blake and M. Isard, \3d position, attitude and shape input using video tracking of hands and lips", In Proc. ACM Siggraph, pages 185-192, 1994.

....The model offers a point to point control and is able to yield more accurately subtle facial deformations such as wrinkles and furrows which are difficult to reproduce with a geometric model. The recent trade in research is mainly recognizing and tracking faces in video images using model based [6, 84] or optical flow based [5, 22] techniques and extracting data from them like facial features which could be used to deriveemotional states of the faces. There are some techniques used for locating human faces in images, and locating facial features in face images. The thesis also mentions briefly ....

A. Blake and M. Isard. 3D position, attitude and shape input using video tracking of hands and lips. In ACM Computer Graphics (Proc. of SIGGRAPH '94), volume 28, pages 185--192, 1994.

....combination of example shapes p i ;std # where the shape of the ith example image, y p i ;std ,isthe 2D warping used to geometrically normalize the image in the off line preparation step. This technique for modeling shape is similar to the work of Cootes, et al. 17] Blake and Isard [12], Baumberg and Hogg [4] and Jones and Poggio [21] The new shape step would, given i a and reference t a , try to find a set of coefficients ff i that minimizes the squared error of the approximation a (x p i ;std (x) t a : This involves replacing the optical flow ....

Andrew Blake and Michael Isard. 3D position, attitude and shape input using video tracking of hands and lips. In SIGGRAPH '94 Proceedings,pages 185--192, Orlando, FL, 1994.

....process. We distinguish em 3 d hand models and appearance based models. 3 d hand models use articulated structures of the human hand to estimate the hand movements [28, 21] whereas appearance based models directly link the appearance of the hand movements in visual images to specific gestures [2, 12, 32]. 3 d hand model based systems often provide a higher flexibility, due to the estimation of joint angles and a higher precision. Finally, the form of output from the tracking process determines the scope of possible applications. We classify 2 d systems [2] e.g. for controlling 2 d user ....

....images to specific gestures [2, 12, 32] 3 d hand model based systems often provide a higher flexibility, due to the estimation of joint angles and a higher precision. Finally, the form of output from the tracking process determines the scope of possible applications. We classify 2 d systems [2], e.g. for controlling 2 d user interfaces [30] systems working in 3 d by supporting relative 3 d positions [23, 12] and systems which are using stereoscopic vision for most accurate, absolute 3 d positions [28, 32, 33] Obviously, only absolute 3 d position is useful for our application ....

[Article contains additional citation context not shown here]

A. Blake and M. Isard 3D position, attitude and shape input using video tracking of hands and lips In Proc. of SIGGRAPH'94

....are suitable for detection and tracking of other facial features as well. For example, eigenfaces have been successfully applied to the problem of recognition and tracking of lips (termed eigenlips [62] 4. 3 Automatic Lipreading Systems Lip movements of a subject are recorded and analyzed [67, 48, 200, 62, 151, 2, 101, 408, 348]. Parameters defining lip shape (for example, width and height of the lip or lip protrusion) are extracted and the phonemic items associated with particular lip shapes are recognized. The first step is to locate the lips on the image. It can be done manually by placing a window on the mouth ....

A. Blake and M. Isard. 3D position, attitude and shape input using video tracking of hands and lips. Computer Graphics Proceedings, Annual Conference Series, pages 185--192, 1994.

....predictor corrector mechanism for least squares estimation for linear systems. Welch and Bishop [147] and Maybeck [93] both provide detailed discussions and mathematical derivations of the Kalman lter for the interested reader. Kalman ltering can be applied to tracking devices, vision tracking [9], and hybrid tracking systems as well [146] B.2.2 Instrumented Gloves Instrumented gloves measure nger movement through various kinds of sensor technology # . These sensors are embedded in a glove or placed on it, usually on the back of the hand. # The exception to this de nition is the ....

Blake, Andrew, and Michael Isard. 3D Position, Attitude and Shape Input Using Video Tracking of Hands and Lips. In Proceedings of SIGGRAPH'94,ACM Press, 185-192, 1994.

....and texture components, using a 3D model to provide correspondences between example face images. The work of Taylor and coworkers et al. 6, 7, 8, 12] on active shape models is probably the closest to ours. Many other flexible models have been proposed, such as the model of Blake and Issard [4]. 2 Linear models In this section we formally specify the linear object class model and describe the matching algorithm used to analyze a novel image in terms of a flexible model. 2.1 Formal specification To write the linear object class model mathematically, we must first introduce some ....

Andrew Blake and Michael Isard. 3d position, attitude and shape input using video tracking of hands and lips. Computer Graphics Proceedings, pages 185--192, 1994.

....approaches have also been proposed. For instance, Terzopoulos and Waters [25] used snakes to follow lines drawn on the user s face. The snakes drive an intricate face model where muscle and skin are physically simulated. In a similar approach, Blake and Isard employed contours in gesture tracking[4], allowing the user to use hand motion as a three dimensional mouse. Rigid motion supplies the position of the mouse while nonrigid motion supplies button pressing and releasing actions. Other researchers [8, 10] have used optic flow to drive the motion of an anatomically motivated polyhedral ....

A. Blake and M. Isard. 3D position, attitude and shape input using video tracking of hands and lips. Proc. SIGGRAPH, 1994.

....work of Viola [Viola, 1995] on alignment of a model with a novel image by maximization of mutual information suggested the stochastic gradient descent algorithm we use in this paper for the matching stage. Many other flexible models have been proposed. We mention here the model of Blake and Issard [Blake and Isard, 1994] , which is similar in spirit to ours. Finally, a very recent paper by Nastar, Moghaddam and Pentland [Nastar et al. 1996] which was published after the work described in this paper was completed, has intriguing similarities with our work, since it is based on the computation of correspondence ....

.... analysis by synthesis because model images are synthesized and then compared to the input image. Another possible application in the image analysis domain is lip reading. A model of lips can be built from examples. This model can then be used to track lips in a sequence of images (see also [Blake and Isard, 1994]) A mapping from model parameters to phonemes can be learned to read the lips. Figure 22 illustrates a general system for image analysis which first matches a 28 linear object class model to a novel image and then maps the coefficients of the model to some higher level parameters by using a ....

Andrew Blake and Michael Isard. 3d position, attitude and shape input using video tracking of hands and lips. Computer Graphics Proceedings, pages 185--192, 1994.

....interactively set a number of values to the desired FAPs to obtain the desired expression. As described in subsection 2.2 these FAPs are generally composed defined manually using 3D animation softwares. They can also be extracted more automatically with the help of external devices such as cameras [6,7,8], or 3D scanners [9,10] The low level layer is the description of the location of the FAP points on the face mesh in a neutral position. Among the high level actions (the only layer visible to the user) one can distinguish three different types of actions: basic emotions, visemes and ....

A.Blake, and M. Isard. 3D position, attitude and shape input using video tracking of hands and lips, Computer Graphics (Proc. SIGGRAPH 94) , 28:185192, July 1994.

....are suitable for detection and tracking of other facial features as well. For example, eigenfaces have been successfully applied to the problem of recognition and tracking of lips (termed eigenlips [49] 4. 3 Automatic Lipreading Systems Lip movements of a subject are recorded and analyzed [54, 40, 161, 49, 125, 2, 85, 346, 291]. Parameters defining lip shape (for example, width and height of the lip or lip protrusion) are extracted and the phonemic items associated with particular lip shapes are recognized. The first step is to locate the lips on the image. It can be done manually by placing a window on the mouth ....

A. Blake and M. Isard. 3D position, attitude and shape input using video tracking of hands and lips. Computer Graphics Proceedings, Annual Conference Series, pages 185--192, 1994.

....hands. The marking free systems of [12, 21] can recognize specific finger or pointing events, but not general gestures. Employing special hardware or off line learning, several researchers have developed successful systems to recognize general hand gestures [22, 14, 6, 7, 20] Blake and Isard [4] have developed a fast contour based tracker which they applied to hands, but the discrimination of different hand poses is limited. The real time hand gesture recognition systems we are aware of require special hardware or lengthy training analysis. 2 Our Approach We seek a simple and fast ....

A. Blake and M. Isard. 3D position, attitude and shape input using video tracking of hands and lips. In Proceedings of SIGGRAPH 94, pages 185--192, 1994. In Computer Graphics, Annual Conference Series.

....hand gesture was found to be somewhat tiring for extended viewing. An improvement may be to maintain the open hand as a trigger gesture, but allow a more restful command gesture, once the trigger gesture has been detected and the hand located. The contour tracking algorithms of Blake and Isard [3] may be useful for such commands. Multiple templates are useful for robust operation of the prototype, although too many templates makes false detection of the trigger more possible. Further development has to be undertaken to determine whether this simple correlation based image processing could ....

A. Blake and M. Isard. 3D position, attitude and shape input using video tracking of hands and lips. In Proceedings of SIGGRAPH 94, pages 185--192, 1994. In Computer Graphics, Annual Conference Series.

No context found.

A. Blake and M. Isard. 3D position, attitude and shape input using video tracking of hands and lips. In Proc. Siggraph, pages 185--192, 1994.

No context found.

A. Blake and M. Isard. 3D position, attitude and shape input using video tracking of hands and lips. In Proc. Siggraph, pages 185--192, 1994.

.... more successful systems which recover continuous parameters (rather than recognising gestures from a discrete vocabulary ) some use a stereo rig (e.g. 5, 20] some are not real time (e.g. 3, 9] while others do not appear to have sufficient accuracy for the applications envisaged here (e.g. [1, 2, 8, 11, 12]) Of these, 1, 11] are the closest to our system in terms of the method used. In both cases, the tracking is good enough to permit navigation through a virtual environment, but not for the fine adjustment of interactive visual tools (e.g. drawing at pixel accuracy) 2 Partitioned sampling and ....

A. Blake and M.A. Isard. 3D position, attitude and shape input using video tracking of hands and lips. In Proc. Siggraph, pages 185--192. ACM, 1994.

....performance. 1 Introduction The use of Kalman filters [1] to track the motion of objects in real time is now a standard weapon in the arsenal of Computer Vision [14, 11, 8, 10, 16, 3] A crucial consideration for effective real time performance is that some form of dynamical model be identified [13, 5, 2] and used as a predictor. In many cases, available models are deterministic based on ordinary differential equations. However, to be usable in a Kalman filtering framework it is crucial that the model contain both deterministic and stochastic components stochastic differential equations. ....

....are deterministic based on ordinary differential equations. However, to be usable in a Kalman filtering framework it is crucial that the model contain both deterministic and stochastic components stochastic differential equations. Such models can be learned effectively from training data [9, 5]. In this paper we develop two significant elaborations for stochastic dynamical models. The first concerns modelling object classes for objects in motion. The second addresses the efficient modelling of couplings between tracked objects. 1.1 Shape and Motion Variability The first problem ....

[Article contains additional citation context not shown here]

A. Blake and M.A. Isard. 3D position, attitude and shape input using video tracking of hands and lips. In Proc. Siggraph, pages 185--192. ACM, 1994.

....the extraction from image sequences of the first four target statistics listed above. 3 Real Time tracking In this section we present the method used to track cars and extract motion parameters. Blake, et al. presented a real time tracker in [2, 3] and an extended version of this in [4]. We use a version of this tracker on a SUN IPX with a Datacell s2200 image capture board. The tracker is based on work originally done by Kass, et al. 8] and uses a B spline representation of a curve. By using a dynamical model, and a restricted search space, real time performance has been ....

.... template ( X; Y) W = 1 0 X 0 0 Y 0 1 0 Y X 0 and M = Theta W T HW Gamma1 W T H where H is the metric matrix converting measurements in control point space to Euclidean measurements (see [3] 0 and 1 are N vectors of zeroes and ones, respectively details are given in [3, 4] (where extensions of the motion model to non affine and nonrigid motion are also given) In this notation we also have W Q = X; Y) T describing the template, which represents the template as Q = 0; 0; 1; 1; 0; 0) T . The affine constraint is satisfied locally and does not restrict ....

[Article contains additional citation context not shown here]

A. Blake & M. Isard. 3D position, attitude and shape input using video tracking of hands and lips. In Siggraph, 1994.

....for those densities. However, it is obviously more satisfactory to measure the actual densities or estimate them from data sequences (x 1 ; x 2 ; Algorithms to do this assuming Gaussian densities are known in the control theory literature [13] and have been applied in computer vision [6, 7, 4]. 1.2 Sampling methods A standard problem in statistical pattern recognition is to find an object parameterised as x with prior p(x) using data z from a single image. This is a simplified, static form of the image sequence problem addressed in this paper. In order to estimate x from z, some ....

....established both for the dynamics of the object and for the measurement process. As mentioned earlier, the parameters x denote a linear transformation of a B spline curve, either an affine deformation, or some non rigid motion. The dynamical model and learning algorithm follow established methods [6, 7]. The model is a stochastic differential equation which, in discrete time, is x t 1 = Ax t B t (8) where A defines the deterministic component of the model and t is a vector of independent standard normal random variables scaled by B so that BB T is the Iterate At time step t 1, ....

[Article contains additional citation context not shown here]

A. Blake and M.A. Isard. 3D position, attitude and shape input using video tracking of hands and lips. In Proc. Siggraph, 185--192. ACM, 1994.

No context found.

A. Blake and M. Isard. 3D position, attitude and shape input using video tracking of hands and lips. In Proceedings of SIGGRAPH 94, pages 185#192, 1994.

No context found.

A. Blake and M. Isard. 3D position, attitude and shape input using video tracking of hands and lips. In Proceedings of SIGGRAPH 94, pages 185--192, 1994. In Computer Graphics, Annual Conference Series.

No context found.

A. Blake and M. Isard. 3D position, attitude and shape input using video tracking of hands and lips. In Proceedings of SIGGRAPH 94, pages 185--192, 1994.

No context found.

Blake, Andrew, and Michael Isard. 3D Position, Attitude and Shape Input Using Video Tracking of Hands and Lips. In Proceedings of SIGGRAPH'94, ACM Press, 185-192, 1994.

No context found.

A. Blake and M. Isard. 3d position, attitude and shape input using video tracking of hands and lips. In Proc. ACM Siggraph, pages 185--192, 1994.

No context found.

A. Blake and M. Isard 3D position, attitude and shape input using video tracking of hands and lips In Proc. of SIGGRAPH'94

First 50 documents

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC