M. Black and A. Jepson, Recognizing temporal trajectories using the condensation algorithm, in Proc. Third IEEE International Conference on Automatic Face and Gesture Recognition, 1998, pp. 16--21.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....supervision Essa [32] Phase Space Constraints Invariance to speed Supervised learning Campbell, Motion Trajectories Joint changes in body Bobick[38] Angles from 3D Model) extension Condensation Auto. extraction Requires trackable Black of motion trajectory icons supervised et al. [20], initialization Isard [71] Recognition of Oscillatory Prediction module Limited gesture Cohen Motion using Position Vector allows real time vocabulary et al. 41] Dynamic System Models recognition w # 85 accuracy Multi class, Multi variate Feature set is adaptive; Segmentation against ....

....hand motions, this same virtue is also a handicap, as dynamic programming behaves poorly when non spatial features are supplied. In addition, dynamic time warping must be applied at every time instant since accurate segmentation at motion boundaries can be dicult without supervision. Black et al. [19, 20] recently introduced a method based on the Condensation algorithm [71, 72] to recognize gestures and human motion. Human motions are modeled as temporal trajectories of estimated parameters over time. Condensation is used to incrementally matchhuman trajectory models to multi variate input data. ....

M. Black and A.D. Jepson, \Recognizing temporal trajectories using the condensation algorithm," Proceedings Third IEEE International Conference on Automatic Face and Gesture Recognition, pp.16-21 Nara, Japan, April 14-16, 1998.

....(deterministic) module. Walter et al. 13,14] show that incorporating current observation information into this scheme gives improved classification results for human trajectory and gesture classification applications. Again, object tracking is performed by a separate module. Black and Jepson [15] use a similar scheme with multiple temporal models as a combined tracker and classifier to analyse an augmented whiteboard. 2.1.2. Overcoming the curse of dimensionality using current time priors Particle filters work well when the conditional densities f y ) are reasonably flat; however, ....

....to enumerating every possible hidden state sequence and summing the probabilities that they give rise to the observed sequence (the shortcut is described in full in Ref. 18] An alternative online approximation to this is possible using a particle filter, this is described by Black and Jepson [15]. Section 6 describes our application of this method within the context of this paper. Problem 2 (recovery of hidden states) is usually tackled by the Viterbi algorithm [19] also described in Ref. 18] This is again a shortcut to evaluating all possible hidden state sequences. Within the ....

[Article contains additional citation context not shown here]

M. Black, A. Jepson, Recognizing temporal trajectories using the condensation algorithm, Proceedings of the International Conference on Automatic Face and Gesture Recognition, 1998, pp. 16 -- 21.

....algorithm proposed by Isard and Blake [6, 7] sug gests a more flexible framework to HMMs. Instead of modelling observation probabilities conditional to a finite set of states, they are continuously propagated over time. For gesture recognition, CONDENSATION has been adopted by Black and Jepson [1]. The model performs fix sized local linear template matching weighted by the conditional observation densities propagated according to CONDENSATION therefore allowing for a global nonlinear time scaling. Unfortunately, the model does not consider measurements covariance (treated independently) ....

....For modelling a temporal structure, let us define a state vector at time as T #Q W X R given by the hidden Markov state V of a model W X . Notice that this HMM based state vector implicitly encodes the information on both phase and temporal scaling used by Black and Jepson [1]. In general, at any time , a temporal structure is fully described by its state history V# TA T 2 T V , its current observation B V and its observation history over time Figure 3. Learning the spatio temporal structure of drawing figure 5 using HMM and EM clustering. The ....

[Article contains additional citation context not shown here]

M. Black and A. Jepson. Recognizing temporal trajectories using the condensation algorithm. In IEEE Conf. on Face & Gesture Recognition, pages 16--21, Nara, Japan, 1998.

....the tracking over time. An algorithm based on the sampling concept for simultaneous tracking and verification has been presented in [22] Modelswitching random sampling trackers have been presented, where the switching occurs between hand tailored dynamical systems [23] and temporal trajectories [24]. An interesting approach for detecting and tracking learned human motion has been presented in [25] where the detection and tracking is performed by calculating a best global labeling of point features using a learned triangular decomposition of the human body. The concept of learning parametric ....

M. Black and A. Jepson, Recognizing temporal trajectories using the condensation algorithm, in Proc. Third IEEE International Conference on Automatic Face and Gesture Recognition, 1998, pp. 1641.

....each feature trajectory is split into sub trajectories and recognition is achieved by maximizing the probability of it being a particular gesture in the eigenspaces of each these sub trajectories. An incremental recognition strategy that is an extension of the condensation algorithm is proposed in [8] to recognize gestures based on the 2D hand trajectory. Gestures are modeled as velocity trajectories and the condensation algorithm is used to incrementally match the gesture models to the input data. A robust hand tracker proposed in [9] is used for tracking the hand in order to extract features ....

M. J. Black and A. D. Jepson, "Recognizing temporal trajectories using the condensation algorithm," in Proc. of IEEE Int. Conf on Automatic Face and Gesture Recognition, 1998.

.... were used for American Sign Language recognition [6] and moments based on image axis projections were used in a hand driven games interface [2] The method used for matching trajectories is similar in some respects to that used by Black and Jepson to drive a Condensation recognition algorithm [1]. The probabilistic method used for treating gestures as sequences of events is similar to a state machine used to parse gestures [7] British Machine Vision Conference 499 2 Gestures for Visually Mediated Interaction In order to illustrate the approach, let us restrict our attention to a set of ....

M. J. Black and A. D. Jepson. Recognizing temporal trajectories using the condensation algorithm. In Proc. 3rd IEEE Int. Conf. on Automatic Face and Gesture Recognition, pages 16--21, Nara, Japan, 1998.

....(CONDENSATION) algorithm was proposed [5, 6] Instead of modelling observation probabilities conditional to a finite set of discrete states, a set of probabilities for different models is continuously propagated over time. For gesture recognition, CONDENSATION has been adopted by Black and Jepson [1]. Unfortunately, the model does not consider measurement covariance therefore is sensitive to noise. It also does not use any prior knowledge on both the state distribution and the observations of a structure. Consequently, a very large number of density samples (over thousands) with localised ....

....knowledge on the structures of behaviours, the state densities p(s t js t;1 ) can only be poorly guessed as the previous estimation plus arbitrary noise. As a result full estimation of the prior p(s t jO t;1 ) can only be obtained through propagation of a very large number (thousands) of samples [1]. This is both expensive and sensitive to outliers. A potential solution to this problem can be derived from Hidden Markov Models (HMMs) A HMM is defined by a number of discrete states q 2 fq 1 #q 2 #: q N g, with probabilistic transitions between states and observation probabilities for each ....

M.J. Black and A.D. Jepson. Recognizing temporal trajectories using the condensation algorithm. In IEEE Conference on Face & Gesture Recognition, pages 16--21, Japan, 1998.

....distribution for the first frame. Condensation has also been applied to the field of temporal modelling (prediction and classification) of objects. Johnson and Hogg [9] use condensation to propagate multiple prediction hypothesis for pedestrian trajectory classification. Black and Jepson [2] use a similar scheme with multiple temporal models as a combined tracker and classifier to analyse an augmented whiteboard. Walter et al. use Black and Jepson s method with continuous (non discrete) Hidden Markov Models to classify the trajectories of people in an office scene. 2.2 Hidden ....

....independent, however component separation can still yield improved search efficiency. 6 Combined Tracking and Behaviour Analysis using Multiple CHMMs In section 2.1 it was described how the Condensation algorithm has wider application than simple object tracking. In particular Black and Jepson [2] use Condensation to track and classify the trajectories of a coloured whiteboard marker. In their scheme multiple trajectory models are used and the gesture classified as one of six actions to be performed. In our scheme we use the combined tracker and classifier paradigm of Black and Jepson ....

Michael Black and Allan Jepson. Recognizing temporal trajectories using the condensation algorithm. In Proc. Initernational Conference on Automatic Face and Gesture Recognition, pages 16--21, 1998.

.... were used for American Sign Language recognition [5] and moments based on image axis projections were used in a handdriven games interface [2] The method used for matching trajectories is similar in some respects to that used by Black and Jepson to drive a Condensation recognition algorithm [1]. The probabilistic method used for treating gestures as sequences of events is similar to a state machine used to parse gestures [6] 2 Gestures for Visually Mediated Interaction In order to illustrate the approach, let us restrict our attention to a set of four gestures. These gestures have ....

M. J. Black and A. D. Jepson. Recognizing temporal trajectories using the condensation algorithm. In Proc. 3nd IEEE Int. Conf. on Automatic Face and

....as the occluded material appears along the trailing edge of the occluding surface. Additionally, writing actions are typically detected as waving gestures. More sophisticated temporal models of gestures will be needed to distinguish between events such as pointing and writing (see for example [4], 10] Finally, a broader user study is called for. A wider sample of presentations will likely reveal distinct presentation types with their own afforances suggesting ways of modifying and expanding the gestures that need to be recognized. Additionally, we have currently not undertaken a user ....

M. J. Black and A. D. Jepson, Recognizing temporal trajectories using the Condensation algorithm, Int. Conf. on Automatic Face and Gesture Recognition, Nara, Japan, pp. 16--21, April 1998.

....as the occluded material appears along the trailing edge of the occluding surface. Additionally, writing actions are typically detected as waving gestures. More sophisticated temporal models of gestures will be needed to distinguish between events such as pointing and writing (see, for example, [4], 10] Finally, a broader user study is called for. A wider sample of presentations will likely reveal distinct presentation types with their own affordances suggesting ways of modifying and expanding the gestures that need to be recognized. Additionally, we have currently not undertaken a user ....

M. J. Black and A. D. Jepson, "Recognizing temporal trajectories using the condensation algorithm," in Int. Conf. Automat. Face and Gesture Recognition, Nara, Japan, Apr. 1998, pp. 16--21.

No context found.

M. Black and A. Jepson, Recognizing temporal trajectories using the condensation algorithm, in Proc. Third IEEE International Conference on Automatic Face and Gesture Recognition, 1998, pp. 16--21.

No context found.

M. Black and A. Jepson. Recognizing temporal trajectories using the condensation algorithm. In Automatic Face and Gesture Recognition, pages 16--21, 1998.

No context found.

M. Black and A. Jepson. Recognizing temporal trajectories using the condensation algorithm. In Face and Gesture Recognition, pages 16--21, 1998.

No context found.

M. Black and A. Jepson. Recognizing temporal trajectories using the condensation algorithm. In Face and Gesture Recognition, pages 16--21, 1998.

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