J. Triesch and C. von de Malsburg, "Robust classification of hand postures against complex background," in Proc. Int'l Conf. on Automatic Face and Gesture Recognition, 1996, pp. 170--175.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....of these makes a gesture. Similarly, Maggioni and Kmmerer [54] described the GestureComputer, which recognized both hand gestures and head movements. Other systems which recognize hand postures amidst complex visual backgrounds are reported by Weng and Cui [55] and Triesch and von der Malsburg [56]. There has been a lot of interest in creating devices to automatically interpret various sign languages to aid the deaf community. One of the first to use computer vision without requiring the user to wear anything special was built by Starner [57] who used HMMs to recognize a limited vocabulary ....

J. Triesch and C. von der Malsburg, "Robust classification of hand postures against complex backgrounds," Proc. Second International Conference on Automatic Face and Gesture Recognition, Killington, VT, Oct. 1996.

.... along the works by [3] who extracted peaks from a Laplacian pyramid of an image and linked them into a tree structure with respect to resolution, 12] who constructed scale space primal sketch with an explicit encoding of blob like structures in scale space as well as the relations between these, [20] who used elastic graphs to represent hands in different postures with local jets of Gabor filters computed at each vertex, 17] who detected maxima in a multi scale wavelet transform. The use of chromaticity as a primary cue for detecting skin coloured regions was first proposed by [5] Our ....

J. Triesch and C. von der Malsburg. Robust classification of hand postures against complex background. In Proc. Int. Conf. on Face and Gesture Recognition, pages 170--175, Killington, Vermont, 1996.

....of different scale matched to different images) 5.2 Matching Table 2 and 3 show the results for two matching tasks, the localization of faces and hand postures. For both tasks matching within the approach described in this chapter is compared to the matching with bunch graphs as described in [46, 20, 43]. The extremely difficult face test set contains 120 frontal faces with uncontrolled illumination and mostly inhomogeneous background. Size variation of the faces is between 15 and 100 pixel (Figure 17 shows some examples of matches and mismatches on this data set) The first row gives the ....

....requires only 4.9 seconds instead of 17 seconds. The simulations corresponding to the third row were performed with representations extracted from only one image. Performance decreases to 80 (set 1) 52 (set 2) and 52 (set 3) The performance with the bunch graph approach as described in [43] is given in the fourth row. For test set 1 and 2 performance is comparable to ORASSYLL (in case of set 2 even slightly better) For set 3 performance is significantly worse compared to ORASSYLL. In [18] simulations with other objects are performed to investigate the influence of variation of ....

[Article contains additional citation context not shown here]

J. Triesch and C. von der Malsburg. Robust classification of hand postures against complex background. Proceedings of the Second International Workshop on Automatic Face- and Gesture recognition, Vermont, pages 170--175, 1996.

....recognition and finger tracking. Interestingly there are many different approaches to this problem with no single dominating method. The basic techniques include color segmentation [8] 12] infrared segmentation [13] blob models [6] contours [13] 9] 15] correlation [4] 11] and wavelets [17]. Typical sample applications are finger driven mice [11] 12] finger driven drawing applications [4] 6] 9] bare hand game control [5] 15] and bare hand television control [5] Most authors use some kind of restriction, to simplify the computer vision process: Non real time ....

.... sample applications are finger driven mice [11] 12] finger driven drawing applications [4] 6] 9] bare hand game control [5] 15] and bare hand television control [5] Most authors use some kind of restriction, to simplify the computer vision process: Non real time calculations [17] . Colored gloves [8] Expensive hardware requirements (e.g. 3D camera or infraredcamera) 13] 14] Restrictive background conditions [15] 1 Every tracker looses track from time to time (e.g. due to occlusion) and has to be restarted. By restarting the system at each frame, many ....

Triesch, J. and Malsburg, C. Robust Classification of Hand Postures Against Complex Background, International Conference On Automatic Face and Gesture Recognition, Killington, 1996.

....devices, a classification of hand postures is often enough in many other applications such as commands switching. Since the appearances are much di#erent among di#erent hand postures and these di#erences are not large among di#erent people, an alternative approach is appearance based approach [5, 6, 14, 19], in which classifiers are learned for a set of image samples. Although it is easier for the appearance based approach to achieve user independence than model based approach, there are two major di#culties of this approach: automatic feature selection and training data collection. Although there ....

....samples. Although it is easier for the appearance based approach to achieve user independence than model based approach, there are two major di#culties of this approach: automatic feature selection and training data collection. Although there have been many discussions about feature extraction [19, 14, 13] and selection [5, 6] little has been addressed on the training data. The generalization of many current methods have to largely depend on their training data sets. In general, good generalization requires a large and representative labeled training data set. However, to manually label a large ....

J. Triesch, C. Malsburg, "Robust Classification of Hand Postures Against Complex Background", Int'l Conf. On Automatic Face and Gesture Recognition, 1996

....hand sign and pose recognition from still imagery. 142] described a general framework for learning based hand sign recognition. Discriminant analysis was used to automatically select the most discriminating features and good recognition results were obtained for 28 different static hand signs. [143] applied an elastic graph matching based approach to gesture recognition. 144] describes the use of hand gesture analysis in combination with speech recognition in a bi modal interface for controlling a 3D display. For a review of hand gesture recognition techniques, see [145] for more detailed ....

J. Triesch and C.v.d. Malsburg, "Robust Classification of Hand Posture against Complex Background," in Proceedings, International Conference on Automatic Face and Gesture Recognition, pp. 170-175, 1996.

....the goal of our field in making computers deal with visual information as visual events occur in human s very complex environment. A lot of challenging work remains to be done toward this goal. Independent of the work presented here, which is completed in Nov. 1995 [11] Triesch von der Malsburg [33] recently used their jets to identify 10 static hand poses from complex background, although segmetation issue was not explicitly addressed in their work. In the current implementation, the fixations are generated mechanically. The number of fixations and the positions of fixations are the same ....

J. Triesch and C. von der Malsburg, "Robust classification of hand postures against complex backgrounds," in Proc. Int'l Conf. Automatic Face and Gesture Recognition, Killington, Vermont, pp. 170-175, Oct. 14-16, 1996.

....the vertices from a sparse grid placed over the object of interest. The vertices of the grid are unlikely to fall over the most distinctive points, so when searching for the grid in a similar image only the vertices which happened to lie near distinctive points are found accurately. Triesch et al. [13] tackles this problem by extracting the vectors from the vertices of a graph which is placed manually over the object. The vertices are chosen to coincide with heavily textured positions (ie distinctive points) Our approach improves on this by providing an automatic, principled way of locating ....

J. Triesh and C. von der Malsburg. Robust classification of hand postures against complex backgrounds. In Automatic Face and Gesture Recognition, pages 170--175, Los Alamitos, California, October 1996. IEEE Computer Society Press.

....at the vertices from a sparse grid placed over the object of interest. The vertices of the grid are unlikely to fall over the most salient points, so when searching for the grid in a similar image only the vertices which happened to lie near salient points are found accurately. Triesch et al. [15] tackle this problem by extracting the vectors from the vertices of a graph which is placed manually over the object. The vertices are chosen to coincide with heavily textured positions (ie salient points) Our approach improves on this by providing an automatic, principled way of locating these ....

J. Triesh and C. von der Malsburg. Robust classification of hand postures against complex backgrounds. In Automatic Face and Gesture Recognition, pages 170--175, Los Alamitos, California, Oct. 1996. IEEE Computer Society Press.

....at the vertices from a sparse grid placed over the object of interest. The vertices of the grid are unlikely to fall over the most salient points, so when searching for the grid in a similar image only the vertices which happened to lie near salient points are found accurately. Triesch et al. [15] tackles this problem by extracting the vectors from the vertices of a graph which is placed manually over the object. The vertices are chosen to coincide with heavily textured positions (ie salient points) Our approach improves on this by providing an automatic, principled way of locating these ....

J. Triesh and C. von der Malsburg. Robust classification of hand postures against complex backgrounds. In Automatic Face and Gesture Recognition, pages 170--175, Los Alamitos, California, Oct. 1996. IEEE Computer Society Press.

.... the specific application to qualitative hand tracking that will be addressed in this paper, more detailed hand models have been presented by (Kuch Huang 1995, Heap Hogg 1996, Yasumuro et al. 1999) Related graph like representations for hand tracking and face tracking have been presented by (Triesch von der Malsburg 1996, Mauerer von der Malsburg 1996) 3 Image Features and Qualitative Feature Relations We are interested in representing objects which can give rise to a rich variety of image features of di#erent types and at di#erent scales. Generically, these image features can be (i) zero dimensional ....

....objects having a large number of features at the same scale level in the hierarchy, we can see several advantages in extend the proposed feature hierarchy by also defining such inter feature relations at the same level. Such relations could be inspired by the works on labeled feature graphs by (Triesch von der Malsburg 1996). Concerning the determination of the qualitative relations between the features in the hierarchy, it would be interesting to explore a framework for learning the relations from training examples. In such a framework and under the assumption that our scheme for feature tracking can register ....

Triesch, J. & von der Malsburg, C. (1996), Robust classification of hand postures against complex background, in `Int. Conf. on Automatic Face and Gesture Recognition', Killington, Vermont, pp. 170--175.

....robot must meet at least two requirements. First, it must be person independent, i.e. the robot must understand commands given by di#erent persons. Second, it must be robust with respect to all the variation and background noise present in natural environments. The interface we have developed [32, 31] allows the operator to transmit commands to GripSee by performing hand gestures, e.g. by pointing at an object in a specific manner in order to have the robot pick up the object in a specific way. It consists of two agents. The first one tracks the operator s hand, the second one performs a ....

.... to Gabor based wavelets, while the edges (in the graph theoretical sense) contain geometrical information (figure 4) In previous work we have demonstrated that elastic graph matching can be successfully applied to the person independent recognition of hand postures in front of complex backgrounds [32]. The system presented there was optimized for the robot application by using graphs with fewer and sparser nodes, as well as by reducing the number of allowed hand postures from ten to six (see table 1 for results for both systems) Higher performance, which is necessary for truly reliable ....

[Article contains additional citation context not shown here]

Jochen Triesch and Christoph von der Malsburg. Robust classification of hand postures against complex backgrounds. In Proceedings of the Second International Conference on Automatic Face and Gesture Recognition, pages 170--175. IEEE Computer Society Press, 1996.

....9:5 seconds and recognition rate for the first set was 93 (first row) The simulations corresponding to the second row were performed with representations extracted by one shot learning. The performance is still remarkably high (80 ) The performance with the bunch graph approach as described in [8] is given in the third row. Results for the test set with uncontrolled background and illumination is shown in row 4 6. For the first test set performance within the bunch graph approach [9, 8] is comparable to ORASSYLL. For the second and more difficult, set performance of ORASSYLL is ....

....performance is still remarkably high (80 ) The performance with the bunch graph approach as described in [8] is given in the third row. Results for the test set with uncontrolled background and illumination is shown in row 4 6. For the first test set performance within the bunch graph approach [9, 8] is comparable to ORASSYLL. For the second and more difficult, set performance of ORASSYLL is significantly better. 6 Comparison with the Jet based System ORASSYLL has been heavily influenced by an older and well known vision system [4, 9, 8] in the following called jet based system, and has ....

[Article contains additional citation context not shown here]

J. Triesch and C. von der Malsburg. Robust classification of hand postures against complex background. Proceedings of the Second International Workshop on Automatic Face- and Gesture recognition, Vermont, pages 170--175, 1996.

.... gloves glove colors as ellipse addressed Lanitis Special None Not Hand tracking O(n) 5 et al. 95 [44] shapes mentioned with parameters Kjeldsen Special None No skin Color histogram O(n) 5 Kender 95 [40] shapes tone and thresholding Triesch Speical None Moderately Labeled O(n) 10 Malsburg [82] shapes complex graphs Cui ASL None Arbitrarily SHOSLIF with O(log(n) 28 Weng 96 [19] words complex visual attention algorithm that uses a systematic tree structure. It is the only system that has been extensively tested for both face recognition and object recognition. 2.7 SHOSLIF M: Motion ....

....[18] 19] is unique in the following sense: 1. The capability to segment a detailed hand (a complex articulated object) from a very complex background as shown in Fig. 7. The Work of Weng at al. is the only one whose recognition result is completely independent of the background. Malsburg at al. [82] requires that the background covered by local views do not affect the local matching significantly. With this capability, we can significantly reduce the constraint on what kind of clothes that the signer can wear. 2. The logarithmic sign retrieval time complexity O(log(n) as indicated in ....

J. Triesch and C. von der Malsburg. Robust classification of hand posture against complex background. In Proc. Int'l Conf. on Automatic Face- and Gesture-Recognition, pages 170--175, Killington, Vermont, Oct. 1996.

....robot must meet at least two requirements. First, it must be person independent, i.e. the robot must understand commands given by different persons. Second, it must be robust with respect to all the variation and background noise present in natural environments. The interface we have developed [32, 31] allows the operator to transmit commands to GripSee by performing hand gestures, e.g. by pointing at an object in a specific manner in order to have the robot pick up the object in a specific way. It consists of two agents. The first one tracks the operator s hand, the second one performs a ....

.... Gabor based wavelets, while the edges (in the graph theoretical sense) contain geometrical information (figure 4) In previous work we have demonstrated that elastic graph matching can be successfully applied to the person independent recognition of hand postures in front of complex backgrounds [32]. The system presented there was optimized for the robot application by using graphs with fewer and sparser nodes, as well as by reducing the num ber of allowed hand postures from ten to six (see table i for results for both systems) Higher per formance, which is necessary for truly reliable ....

[Article contains additional citation context not shown here]

Jochen Triesch and Christoph von der Malsburg. Robust classification of hand postures against complex backgrounds. In Proceedings of the Second International Conference on Automatic Face and Gesture Recognition, pages 170-175. IEEE Computer Society Press, 1996.

....signing hand. Our goal is to develop a method for accurate estimation of the posture of the gesturing hand. For real world applications it is important that the recognition works despite complex cluttered backgrounds, where there may not be an easy way of segmenting the hand from the background [6]. This is arguably the hardest problem in vision based gesture recognition [7] A system requiring a simple uniform or static background [8] may not be flexible enough for many real world applications. Nevertheless, hardly any work in the literature has dealt with this problem. Researches mostly ....

J. Triesch, C. von der Malsburg, Robust classification of hand postures against complex backgrounds, Proc. Second Int. Conf. Autom. Face Gesture Recogn., IEEE Comput. Soc. Press (1996) 170 -- 175.

....track and recognize a person walking by a camera in real time. On line facial expression recognition [Hong et al. 1997] opens vistas on better human machine communication, for instance for video games, tele conferencing and computer based training systems. Recognition of facial and hand gestures [Triesch and von der Malsburg, 1996] can be used to control machines more conveniently. Fully immersive tele conferencing requires the creation of a display that renders remote participants with correct direction of gaze independent of their spatial relation to the camera, and the creation of a realistic three dimensional sound ....

Triesch, J., and von der Malsburg, C. (1996): Robust Classification of Hand Postures against Complex Backgrounds. Proc. Intl. Workshop on Automatic Face- and Gesture- Recognition, Vermont, 170--175.

....onto an input image. Although the background has large regions of skin color the graph is positioned properly. We have designed a series of systems which have been shown to work in a person independent way and despite varying complex backgrounds. The older systems are described in more detail in [18, 20]. Our previous system [20] only used shape information in the form of responses to Gabor wavelets to the intensity distribution of the image. It performed correctly in four out of five times for six different postures in front of moderately complex backgrounds (see figure 1 right) Recognition ....

J. Triesch and C. v.d. Malsburg. Robust classification of hand postures against complex backgrounds. In Proceedings of the Second International Conference on Automatic Face and Gesture Recognition 1996, Killington, Vermont, USA, October 14-16, 1996.

....by another feedforward net. The complex cells, in turn, can be combined to more complicated feature detectors such as corner detectors [5] They have also proven useful for higher image understanding tasks such as texture classification [6] face recognition [7] 8] and gesture recognition [9]. If the Gabor functions are arranged into a wavelet transform and the sampling is dense enough (see section II D) then the original image can be recovered from the transform values with arbitrary quality (except for the DCvalue) Given the useful properties of the magnitudes of the Gabor ....

Jochen Triesch and Christoph von der Malsburg, "Robust classification of hand postures against complex backgrounds," in Proceedings of the Second International Conference on Automatic Face and Gesture Recognition. 1996, pp. 170--175, IEEE Computer Society Press.

....landmark. e.g. in a bunch graph a left eye of a frontal face is represented as a set of jets extracted from the left eye of frontal faces of different persons. The bunch graph idea is successfully applied to other object recognition problems, e.g. the discrimination of hand gestures [33] and pose estimation [18] In [35] each landmark was described by approximately 70 jets, each containing 40 complex values. Especially to represent contour edges hitting the background a large amount of jets would be necessary to cover all possible combinations of this edge and the different ....

....represent contour edges hitting the background a large amount of jets would be necessary to cover all possible combinations of this edge and the different backgrounds. With our banana approach we can reduce the data needed to represent a landmark to a few banana wavelets. Furthermore in [35] and [33] the creation of an object representation is very time consuming, because for each view of an object an object dependent grid has to be defined and the landmarks has to be positioned manually for the pictures used to create the bunch graphs. First steps towards an automatic generaition of an ....

[Article contains additional citation context not shown here]

J. Triesch and C. von der Malsburg, "Robust Classification Of Hand Postures Against Complex Backgrounds", Proceddings of the second international Conference on Automatic Face and Gesture Recognition, Vermont 1996.

....objects per image) Figure 3: The twelve postures used in the study. The model graphs are created from six example images (3 persons, light and dark background) 3.2 Hand Posture Recognition The second example is the recognition of handpostures (fig. 3) against very complex backgrounds (fig. 4) [3]. All graphs contain 15 nodes and 20 edges. The nodes were placed manually at anatomically significant points on 6 training images for every posture. We extracted jets of three different feature types from every node of a posture s six models (compare fig. 1) ffl Gabor Features: Responses of ....

....a graph for every feature type for every training image of every posture giving a total of 216 graphs. The six graphs of one posture containing features of the same type were fused into a bunchgraph, expressing the variability of the features among different models and variability of backgrounds [4, 3]. The resulting three bunch graphs for each posture were then fused into a single compound graph for each posture. During a matching process we allowed for 15 degree rotation of the model graph s node positions in the image plane as well as 20 rescaling. After that every node is allowed to move ....

J. Triesch and C. von der Malsburg. Robust classification of hand postures against complex backgrounds. In Proceedings of the Second International Conference on Automatic Face and Gesture Recognition 1996, Killington, Vermont, USA, October 14-16, 1996.

No context found.

J. Triesch and C. von de Malsburg, "Robust classification of hand postures against complex background," in Proc. Int'l Conf. on Automatic Face and Gesture Recognition, 1996, pp. 170--175.

No context found.

J. TRIESCH AND C. VON DER MALSBURG, Robust Classification of Hand Postures against Complex Backgrounds, in Proc. IEEE Intl. Conference on Automatic Face and Gesture Recognition, October 1996.

No context found.

J. Triesch and C. von de Malsburg. Robust classification of hand postures against complex background. In Proc. Int'l Conf. On Automatic Face and Gesture Recognition, 1996.

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