Automatic 2D Hand Tracking in Video Sequences

In gesture and sign language video sequences, hand motion tends to be rapid, and hands frequently appear in front of each other or in front of the face. Thus, hand location is often ambiguous, and naive color-based hand tracking is insufficient. To improve tracking accuracy, some methods employ a prediction-update framework, but such methods require careful initialization of model parameters, and tend to drift and lose track in extended sequences. In this paper, a temporal filtering framework for hand tracking is proposed that can initialize and reset itself without human intervention. In each frame, simple features like color and motion residue are exploited to identify multiple candidate hand locations. The temporal filter then uses the Viterbi algorithm to select among the candidates from frame to frame. The resulting tracking system can automatically identify video trajectories of unambiguous hand motion, and detect frames where tracking becomes ambiguous because of occlusions or overlaps. Experiments on video sequences of several hundred frames in duration demonstrate the system's ability to track hands robustly, to detect and handle tracking ambiguities, and to extract the trajectories of unambiguous hand motion.

Skin Color-Based Video Segmentation under Time-Varying Illumination

A novel approach for real-time skin segmentation in video sequences is described. The approach enables reliable skin segmentation despite wide variation in illumination during tracking. An explicit second order Markov model is used to predict evolution of the skin-color (HSV) histogram over time. Histograms are dynamically updated based on feedback from the current segmentation and predictions of the Markov model. The evolution of the skin-color distribution at each frame is parameterized by translation, scaling and rotation in color space. Consequent changes in geometric parameterization of the distribution are propagated by warping and resampling the histogram. The parameters of the discrete-time dynamic Markov model are estimated using Maximum Likelihood Estimation, and also evolve over time. The accuracy of the new dynamic skin color segmentation algorithm is compared to that obtained via a static color model. Segmentation accuracy is evaluated using labeled ground-truth video sequences taken from staged experiments and popular movies. An overall increase in segmentation accuracy of up to 24% is observed in 17 out of 21 test sequences. In all but one case the skin-color classification rates for our system were higher, with background classification rates comparable to those of the static segmentation.

Foreground Object Segmentation from Binocular Stereo Video

Moving cameras are needed for a wide range of applications in robotics, vehicle systems, surveillance, etc. However, many foreground object segmentation methods reported in the literature are unsuitable for such settings; these methods assume that the camera is fixed and the background changes slowly, and are inadequate for segmenting objects in video if there is significant motion of the camera or background. To address this shortcoming, a new method for segmenting foreground objects is proposed that utilizes binocular video. The method is demonstrated in the application of tracking and segmenting people in video who are approximately facing the binocular camera rig. Given a stereo image pair, the system first tries to find faces. Starting at each face, the region containing the person is grown by merging regions from an over-segmented color image. The disparity map is used to guide this merging process. The system has been implemented on a consumer-grade PC, and tested on video sequences of people indoors obtained from a moving camera rig. As can be expected, the proposed method works well in situations where other foreground-background segmentation methods typically fail. We believe that this superior performance is partly due to the use of object detection to guide region merging in disparity/color foreground segmentation, and partly due to the use of disparity information available with a binocular rig, in contrast with most previous methods that assumed monocular sequences.

Estimating 3D Body Pose Using Uncalibrated Cameras

An approach for estimating 3D body pose from multiple, uncalibrated views is proposed. First, a mapping from image features to 2D body joint locations is computed using a statistical framework that yields a set of several body pose hypotheses. The concept of a "virtual camera" is introduced that makes this mapping invariant to translation, image-plane rotation, and scaling of the input. As a consequence, the calibration matrices (intrinsics) of the virtual cameras can be considered completely known, and their poses are known up to a single angular displacement parameter. Given pose hypotheses obtained in the multiple virtual camera views, the recovery of 3D body pose and camera relative orientations is formulated as a stochastic optimization problem. An Expectation-Maximization algorithm is derived that can obtain the locally most likely (self-consistent) combination of body pose hypotheses. Performance of the approach is evaluated with synthetic sequences as well as real video sequences of human motion.

Specialized Mappings Architecture with Applications to Vision-based Estimation of Articulated Body Pose

A fundamental task of vision systems is to infer the state of the world given some form of visual observations. From a computational perspective, this often involves facing an ill-posed problem; e.g., information is lost via projection of the 3D world into a 2D image. Solution of an ill-posed problem requires additional information, usually provided as a model of the underlying process. It is important that the model be both computationally feasible as well as theoretically well-founded. In this thesis, a probabilistic, nonlinear supervised computational learning model is proposed: the Specialized Mappings Architecture (SMA). The SMA framework is demonstrated in a computer vision system that can estimate the articulated pose parameters of a human body or human hands, given images obtained via one or more uncalibrated cameras. The SMA consists of several specialized forward mapping functions that are estimated automatically from training data, and a possibly known feedback function. Each specialized function maps certain domains of the input space (e.g., image features) onto the output space (e.g., articulated body parameters). A probabilistic model for the architecture is first formalized. Solutions to key algorithmic problems are then derived: simultaneous learning of the specialized domains along with the mapping functions, as well as performing inference given inputs and a feedback function. The SMA employs a variant of the Expectation-Maximization algorithm and approximate inference. The approach allows the use of alternative conditional independence assumptions for learning and inference, which are derived from a forward model and a feedback model. Experimental validation of the proposed approach is conducted in the task of estimating articulated body pose from image silhouettes. Accuracy and stability of the SMA framework is tested using artificial data sets, as well as synthetic and real video sequences of human bodies and hands.

Active Hidden Models for Tracking with Kernel Projections

We introduce Active Hidden Models (AHM) that utilize kernel methods traditionally associated with classification. We use AHMs to track deformable objects in video sequences by leveraging kernel projections. We introduce the "subset projection" method which improves the efficiency of our tracking approach by a factor of ten. We successfully tested our method on facial tracking with extreme head movements (including full 180-degree head rotation), facial expressions, and deformable objects. Given a kernel and a set of training observations, we derive unbiased estimates of the accuracy of the AHM tracker. Kernels are generally used in classification methods to make training data linearly separable. We prove that the optimal (minimum variance) tracking kernels are those that make the training observations linearly dependent.

Estimation and Prediction of Evolving Color Distributions for Skin Segmentation Under Varying Illumination

A novel approach for real-time skin segmentation in video sequences is described. The approach enables reliable skin segmentation despite wide variation in illumination during tracking. An explicit second order Markov model is used to predict evolution of the skin color (HSV) histogram over time. Histograms are dynamically updated based on feedback from the current segmentation and based on predictions of the Markov model. The evolution of the skin color distribution at each frame is parameterized by translation, scaling and rotation in color space. Consequent changes in geometric parameterization of the distribution are propagated by warping and re-sampling the histogram. The parameters of the discrete-time dynamic Markov model are estimated using Maximum Likelihood Estimation, and also evolve over time. Quantitative evaluation of the method was conducted on labeled ground-truth video sequences taken from popular movies.

Inferring Body Pose without Tracking Body Parts

A novel approach for estimating articulated body posture and motion from monocular video sequences is proposed. Human pose is defined as the instantaneous two dimensional configuration (i.e., the projection onto the image plane) of a single articulated body in terms of the position of a predetermined set of joints. First, statistical segmentation of the human bodies from the background is performed and low-level visual features are found given the segmented body shape. The goal is to be able to map these, generally low level, visual features to body configurations. The system estimates different mappings, each one with a specific cluster in the visual feature space. Given a set of body motion sequences for training, unsupervised clustering is obtained via the Expectation Maximation algorithm. Then, for each of the clusters, a function is estimated to build the mapping between low-level features to 3D pose. Currently this mapping is modeled by a neural network. Given new visual features, a mapping from each cluster is performed to yield a set of possible poses. From this set, the system selects the most likely pose given the learned probability distribution and the visual feature similarity between hypothesis and input. Performance of the proposed approach is characterized using a new set of known body postures, showing promising results.

Tracking Human Body Pose on a Learned Smooth Space

Particle filtering is a popular method used in systems for tracking human body pose in video. One key difficulty in using particle filtering is caused by the curse of dimensionality: generally a very large number of particles is required to adequately approximate the underlying pose distribution in a high-dimensional state space. Although the number of degrees of freedom in the human body is quite large, in reality, the subset of allowable configurations in state space is generally restricted by human biomechanics, and the trajectories in this allowable subspace tend to be smooth. Therefore, a framework is proposed to learn a low-dimensional representation of the high-dimensional human poses state space. This mapping can be learned using a Gaussian Process Latent Variable Model (GPLVM) framework. One important advantage of the GPLVM framework is that both the mapping to, and mapping from the embedded space are smooth; this facilitates sampling in the low-dimensional space, and samples generated in the low-dimensional embedded space are easily mapped back into the original highdimensional space. Moreover, human body poses that are similar in the original space tend to be mapped close to each other in the embedded space; this property can be exploited when sampling in the embedded space. The proposed framework is tested in tracking 2D human body pose using a Scaled Prismatic Model. Experiments on real life video sequences demonstrate the strength of the approach. In comparison with the Multiple Hypothesis Tracking and the standard Condensation algorithm, the proposed algorithm is able to maintain tracking reliably throughout the long test sequences. It also handles singularity and self occlusion robustly.

Multi-scale 3D Scene Flow from Binocular Stereo Sequences

Scene flow methods estimate the three-dimensional motion field for points in the world, using multi-camera video data. Such methods combine multi-view reconstruction with motion estimation approaches. This paper describes an alternative formulation for dense scene flow estimation that provides convincing results using only two cameras by fusing stereo and optical flow estimation into a single coherent framework. To handle the aperture problems inherent in the estimation task, a multi-scale method along with a novel adaptive smoothing technique is used to gain a regularized solution. This combined approach both preserves discontinuities and prevents over-regularization-two problems commonly associated with basic multi-scale approaches. Internally, the framework generates probability distributions for optical flow and disparity. Taking into account the uncertainty in the intermediate stages allows for more reliable estimation of the 3D scene flow than standard stereo and optical flow methods allow. Experiments with synthetic and real test data demonstrate the effectiveness of the approach.

Handsignals Recognition From Video Using 3D Motion Capture Data

Hand signals are commonly used in applications such as giving instructions to a pilot for airplane take off or direction of a crane operator by a foreman on the ground. A new algorithm for recognizing hand signals from a single camera is proposed. Typically, tracked 2D feature positions of hand signals are matched to 2D training images. In contrast, our approach matches the 2D feature positions to an archive of 3D motion capture sequences. The method avoids explicit reconstruction of the 3D articulated motion from 2D image features. Instead, the matching between the 2D and 3D sequence is done by backprojecting the 3D motion capture data onto 2D. Experiments demonstrate the effectiveness of the approach in an example application: recognizing six classes of basketball referee hand signals in video.

Multi-Scale 3D Scene Flow from Binocular Stereo Sequences

Scene flow methods estimate the three-dimensional motion field for points in the world, using multi-camera video data. Such methods combine multi-view reconstruction with motion estimation. This paper describes an alternative formulation for dense scene flow estimation that provides reliable results using only two cameras by fusing stereo and optical flow estimation into a single coherent framework. Internally, the proposed algorithm generates probability distributions for optical flow and disparity. Taking into account the uncertainty in the intermediate stages allows for more reliable estimation of the 3D scene flow than previous methods allow. To handle the aperture problems inherent in the estimation of optical flow and disparity, a multi-scale method along with a novel region-based technique is used within a regularized solution. This combined approach both preserves discontinuities and prevents over-regularization – two problems commonly associated with the basic multi-scale approaches. Experiments with synthetic and real test data demonstrate the strength of the proposed approach.

An alignment based similarity measure for hand detection in cluttered sign language video

Locating hands in sign language video is challenging due to a number of factors. Hand appearance varies widely across signers due to anthropometric variations and varying levels of signer proficiency. Video can be captured under varying illumination, camera resolutions, and levels of scene clutter, e.g., high-res video captured in a studio vs. low-res video gathered by a web cam in a user’s home. Moreover, the signers’ clothing varies, e.g., skin-toned clothing vs. contrasting clothing, short-sleeved vs. long-sleeved shirts, etc. In this work, the hand detection problem is addressed in an appearance matching framework. The Histogram of Oriented Gradient (HOG) based matching score function is reformulated to allow non-rigid alignment between pairs of images to account for hand shape variation. The resulting alignment score is used within a Support Vector Machine hand/not-hand classifier for hand detection. The new matching score function yields improved performance (in ROC area and hand detection rate) over the Vocabulary Guided Pyramid Match Kernel (VGPMK) and the traditional, rigid HOG distance on American Sign Language video gestured by expert signers. The proposed match score function is computationally less expensive (for training and testing), has fewer parameters and is less sensitive to parameter settings than VGPMK. The proposed detector works well on test sequences from an inexpert signer in a non-studio setting with cluttered background.

Exploiting Phonological Constraints for Handshape Inference in ASL Video

Handshape is a key articulatory parameter in sign language, and thus handshape recognition from signing video is essential for sign recognition and retrieval. Handshape transitions within monomorphemic lexical signs (the largest class of signs in signed languages) are governed by phonological rules. For example, such transitions normally involve either closing or opening of the hand (i.e., to exclusively use either folding or unfolding of the palm and one or more fingers). Furthermore, akin to allophonic variations in spoken languages, both inter- and intra- signer variations in the production of specific handshapes are observed. We propose a Bayesian network formulation to exploit handshape co-occurrence constraints, also utilizing information about allophonic variations to aid in handshape recognition. We propose a fast non-rigid image alignment method to gain improved robustness to handshape appearance variations during computation of observation likelihoods in the Bayesian network. We evaluate our handshape recognition approach on a large dataset of monomorphemic lexical signs. We demonstrate that leveraging linguistic constraints on handshapes results in improved handshape recognition accuracy. As part of the overall project, we are collecting and preparing for dissemination a large corpus (three thousand signs from three native signers) of American Sign Language (ASL) video. The video have been annotated using SignStream® [Neidle et al.] with labels for linguistic information such as glosses, morphological properties and variations, and start/end handshapes associated with each ASL sign.

Extraction and Clustering of Motion Trajectories in Video

A system is described that tracks moving objects in a video dataset so as to extract a representation of the objects' 3D trajectories. The system then finds hierarchical clusters of similar trajectories in the video dataset. Objects' motion trajectories are extracted via an EKF formulation that provides each object's 3D trajectory up to a constant factor. To increase accuracy when occlusions occur, multiple tracking hypotheses are followed. For trajectory-based clustering and retrieval, a modified version of edit distance, called longest common subsequence (LCSS) is employed. Similarities are computed between projections of trajectories on coordinate axes. Trajectories are grouped based, using an agglomerative clustering algorithm. To check the validity of the approach, experiments using real data were performed.