A. Bobick and J. Davis. An appearance-based representation of action. ICPR, 1996.  Home/Search   Document Details and Download   Summary   Related Articles   Check

....with. Second, detailed people detection and tracking is very di#cult. In each frame of the video, quite often several body parts are occluded. Many research initiatives are under way. Most researches have been focusing on detecting a set of predefined activities, such as people walking vs. running[6, 2, 4], certain gesture and interaction[3, 18, 21] or specific events such as someone stealing a suitcase[11, 14] Typically, simple activities are modelled as spatial temporal patterns, and complex activities are represented as a hierarchical combination of simpler activities conditioned on various ....

A. F. Bobick and J. W. Davis. "An appearance-based representation of action." In 13th International Conference on Pattern Recognition, Vienna, Austria, August 1996.

.... amenable to matching by global transforms (such as the linear transformation we consider) Also, this global feature allows recognition based on partial or corrupted data (including missing onset or offset data) The most closely related work to the work reported here is that of Bobick and Davis [7] and Ju et al. 11] both proposed using principal com ponent analysis to model parameters computed from activities but did not demonstrate modeling and recognition of activities. Also, Liet al... 12] proposed a PCA based modeling and recognition approach that exploited entire image sequences of ....

A. Bobick and J. Davis, An appearance-based representation of action, International Conference on Pattern Recognition 1996, 307- 312.

....temporal curves corresponding to joint motion may be derived from biometric studies or learned from 3D motion capture data. In previous work on principal component analysis of motion data, the 3D motion curves corresponding to particular activities had typically to be hand segmented and aligned [1, 7, 8]. By contrast, this paper details an automated method for segmenting the data into individual activities, aligning activities from different examples, modeling the statistical variation in the data, dealing with missing data, enforcing smooth transitions between cycles, and deriving a ....

A. Bobick and J. Davis. An appearance-based representation of action. ICPR, 1996.

....standard methods [10, 9] What makes this problem difficult is the confounding of motion, structure, and imaging geometry such that the appearances of different objects are often indistinguishable from one another. Similar problems are inherent in recognizing particular motions, e.g. gestures [8, 5, 7, 4], where the problem is often made tractable by the dominance of the motion component of the optical flow field. In fact, this is precisely what we wish to avoid in the object recognition context, hence some control of imaging parameters is required to ensure that a reasonable component of the flow ....

A. F. Bobick and J. W. Davis. An appearance-based representation of action. Technical Report 369, MIT Media Lab, February 1996. As submitted to ICPR 96.

....is very important to his taxonomy. Simple and linear for the first case, dynamic time warping for the second, and reasoning about temporal relationships for the last case. Comments: Nice overview paper with a good motion distinction taxonomy. Title: An Appearance based Representation of Action [14] Author(s) A.F. Bobick and J.W. Davis Location: M.I.T. Media Laboratory, USA Year: 1996 Published: International Conference on Pattern Recognition Type: Paper 12 Key words: View based action recognition, motion energy image, PCA, stick figure and optical flow Summary: This work is part of ....

A.F. Bobick and J.W. Davis. An Appearance-based Representation of Action. In International Conference on Pattern Recognition, 1996.

....on a set of parameters, and in [5] coupled HMMs are used for modeling interactions of two mobile parts. In [17, 1] Bayesian Networks are used for recognition tasks. Local representation of motion based on optical flow has been exploited in [14, 15] and view based methods are proposed in [4, 2, 12]. Other approaches are based on principal component analysis [28] parameterization of the motion on joint angles [7] and snake fitting [19] Estimation of motion from stereo [27] and multiple view systems [11] has also been investigated. In [6] a mixed state statistical model for the ....

A. F. Bobick. Appearance-based representation of action. 1996.

.... and Metaxas [81] 1995 Kameda et al. 85] 1995 Leung and Yang [98] 1995 Pentland [122] 1995 Pentland [123] 1995 Tesei et al. 136] 1996 Azarbayejani et al. 9] 1996 Becker and Pentland [14] 1996 Bobick [19] 1996 Bobick and Davis [21] 1996 Cai and Aggarwal [27] 1996 Gavrila and Davis [55] 1996 Ju [77] 1996 Ju et al. 78] 1996 Kakadiaris and Metaxas [82] 1996 Kameda and Minoh [83] 1996 Luc [46] 1996 Moezzi et al. 109] 1996 Turk [138] 1996 Wren et ....

A.F. Bobick and J.W. Davis. An Appearance-based Representation of Action. In International Conference on Pattern Recognition, 1996. 66

....to represent the distribution of its motion. The movement of the centroid characterizes the external forces on an object, while the deformation of the object is computed from the dispersion (the eigenvalues of the covariance matrix) or ratio of lengths of the moment ellipse. Bobick and Davis [6] introduced the Motion Energy Image (MEI) a smoothed description of the cumulative spatial distribution of motion energy in a motion sequence. They match this description of motion against stored models of known actions. Bobick and Davis [7] enhanced the MEI to form a motion history image (MHI) ....

....and diamond, respectively. 40 50 60 70 80 90 100 110 120 0 50 100 150 200 250 300 centroids weighted centroids sequences for each time varying scalar, S i . The next section describes how we compute the frequency and phase for each of these signals. The Motion Energy Image (MEI) [6] is arrived at by: binary threshold of motion displacements computed by thresholding the pixelwise summed squared difference between each image and the first, over an entire sequence. The features characterizing the MEI are a set of the seven Hu moments, which are translation, scale, and ....

Bobick, A. F. and Davis, J. W. An appearance-based representation of action. In Proc. 13th International Conference on Pattern Recognition, 1996.

....data corresponding to particular activities. The variation across subjects is modeled by principal component analysis (PCA) of the curve data. Here the first few principal components capture most of the variation in the training set. This approach has been used for representing 2D image changes [2], optical flow [21] and 3D joint angles [18] A related approach uses vector quantization [12] rather than PCA. The primary use of such detailed models is in tracking using 3D articulated models of people. Given the high dimensionality of the human body, the temporal curves are used to constrain ....

A. Bobick and J. Davis. An appearance-based representation of action. ICPR, 1996.

....vector in suitable basis, as in [17] and the recognition done with HMMs as before, or, as in [4] a spatio temporal representation of the optical flow can be built. In [16] a biological motion pattern is represented by a linear combination of prototypical image sequences. Other methods, such as [7, 3, 5, 36], use a statistical hierarchical approach based on local appearance. Others look at different spatio temporal features [8, 28, 23, 35, 24, 15] In this paper we take a different approach: we do not choose local features, and we do not compute optical flow. Instead, we start from the assumption ....

A. F. Bobick. Appearance-based representation of action. 1996.

....temporal curves corresponding to joint motion may be derived from biometric studies or learned from 3D motion capture data. In previous work on principal component analysis of motion data, the 3D motion curves corresponding to particular activities had typically to be hand segmented and aligned [1, 7, 8]. By contrast, this paper details an automated method for segmenting the data into individual activities, aligning activities from di erent examples, modeling the statistical variation in the data, dealing with missing data, enforcing smooth transitions between cycles, and deriving a probabilistic ....

A. Bobick and J. Davis. An appearance-based representation of action. ICPR, 1996.

....data corresponding to particular activities. The variation across subjects is modeled by principal component analysis (PCA) of the curve data. Here the first few principal components capture most of the variation in the training set. This approach has been used for representing 2D image changes [2], optical flow [21] and 3D joint angles [18] A related approach uses vector quantization [12] rather than PCA. The primary use of such detailed models is in tracking using 3D articulated models of people. Given the high dimensionality of the human body, the temporal curves are used to constrain ....

A. Bobick and J. Davis. An appearance-based representation of action. ICPR, 1996.

....of articulated 3D cylinders [12] viewed under perspective projection. Temporal Models. Temporal models of body limb or joint motion also vary in complexity; they include smooth motion [7] linear dynamical models [18] non linear models learned from training data using dimensionality reduction [3, 16, 23], and probabilistic Hidden Markov Models (HMM s) e.g. 4] In many of these methods, image measurements are rst computed and then the temporal models are applied to either smooth or interpret the results. For example, Leventon and Freeman [16] proposed a Bayesian framework for recovering 3D ....

....of the data are illustrated in Figure 3. From the data, m = 13 example walking cycles from 4 di erent subjects (professional dancers) were segmented manually and scaled to the same length. These cycles are then used to train a walking model using Multivariate Principal Component Analysis (MPCA) [3, 19, 23]. In addition to the joint angles, we model the speed, v i of the torso in the direction of the walking motion i. This speed, v i; at time step in the cycle is v i; k g i; 1 g i; k. The curves corresponding to the speed of the torso and the relative angles of the limbs, l i , ....

A. Bobick and J. Davis. An appearance-based representation of action. ICPR, 1996.

....weighted by u , v and the magnitude of motion. For periodic motion, these moments vary periodically with one frequency and its harmonics but di#er in phase. The phase features are used to recognized individuals while walking. 11] extend their work using statistical learning. Bobick and Davis[6] introduced the the Motion Energy Image (MEI) a smoothed description of the cumulative spatial distribution of motion energy in a motion sequence. They match this description of motion against stored models of known actions. In [7] they enhanced the MEI to form a motion history image (MHI) where ....

Aaron F. Bobick and James W. Davis. An appearance-based representation of action. In Proc. 13th International Conference on Pattern Recognition, 1996.

....motion based on where motion has occurred and what are the temporal characteristics of it. The result of this space time process will collapse the time dimension to a static depiction of the action. These static representations are called Motion Energy Images (MEI) and Motion History Images (MHI) [3, 6]. They are functions of the observed motion properties at the corresponding spatial image location in the image sequence. 4.1 Motion Detection Our input video consists of a sequence of color frames in RGB space. By using a simple linear standard operation, we perform a color space conversion to ....

A. Bobick and J Davis. An appearance-based representation of action. In ICPR, 1996.

....learned. The main problem of this method is the requirement of having stationary objects, and the insufficiency of the representation to discriminate among similar motions. Motion analysis techniques have had the problem that registration of useful, filtered information is a hard labor by itself [9, 4, 19]. Our system provides a functional front end that supports such tasks. 3 Basic Approach A diagram of our approach is illustrated in Fig. 1. The first stage of the algorithm is based on the background subtraction methods of [30] The system is initialized by acquiring statistical measurements of ....

....moving object throughout the tracking sequence, despite changes in scale and position. The resulting translation scale stabilized images of the object are then fed to an action recognition module. Actions are represented in terms of motion energy images (MEI s) and motion history images (MHI s) [4, 9]. AnMEI is a cumulative motion image, and an MHI is a function of the recency of the motion at every pixel. By using stabilized input sequences, it is possible to make the MEI MHI approach invariant to unrestricted 3D translational motion. The stabilized representation is then fed to a ....

[Article contains additional citation context not shown here]

A. Bobick and J Davis. An appearance-based representation of action. ICPR, 1996.

No context found.

A. Bobick and J. Davis. An appearance-based representation of action. ICPR, 1996.

No context found.

A. Bobick and J. Davis. An appearance-based representation of action. ICPR, 1996.

No context found.

A. F. Bobick and J. Davis. An appearance-based representation of action. In Proceedings of the IEEE Conference on Pattern Recognition, pages 307--312, 1996.

No context found.

A. Bobick and J Davis. An appearance-based representation of action. ICPR, 1996.

No context found.

A. Bobick and J. Davis. An appearance-based representation of action. ICPR, 1996, 307-312.

No context found.

A. F. Bobick and J. Davis. An appearance-based representation of action. In Proceedings of the IEEE Conference on Pattern Recognition, pages 307--312, 1996.

No context found.

A. Bobick and J. Davis. An appearance-based representation of action. ICPR, 1996.

No context found.

Bobick and Davis: "An Appearance-based Representation of Action", M.I.T. Media Laboratory Perceptual Computing Section Technical Report, 1995, No.369.

No context found.

A.F. Bobick and J.W. Davis. An Appearance-based Representation of Action. In International Conference on Pattern Recognition, 1996.

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