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.

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.

Scalable Peer-to-Peer Indexing with Constant State

We present a distributed indexing scheme for peer to peer networks. Past work on distributed indexing traded off fast search times with non-constant degree topologies or network-unfriendly behavior such as flooding. In contrast, the scheme we present optimizes all three of these performance measures. That is, we provide logarithmic round searches while maintaining connections to a fixed number of peers and avoiding network flooding. In comparison to the well known scheme Chord, we provide competitive constant factors. Finally, we observe that arbitrary linear speedups are possible and discuss both a general brute force approach and specific economical optimizations.

A Two-Tiered On-Line Server-Side Bandwidth Reservation Framework for the Real-Time Delivery of Multiple Video Streams

The advent of virtualization and cloud computing technologies necessitates the development of effective mechanisms for the estimation and reservation of resources needed by content providers to deliver large numbers of video-on-demand (VOD) streams through the cloud. Unfortunately, capacity planning for the QoS-constrained delivery of a large number of VOD streams is inherently difficult as VBR encoding schemes exhibit significant bandwidth variability. In this paper, we present a novel resource management scheme to make such allocation decisions using a mixture of per-stream reservations and an aggregate reservation, shared across all streams to accommodate peak demands. The shared reservation provides capacity slack that enables statistical multiplexing of peak rates, while assuring analytically bounded frame-drop probabilities, which can be adjusted by trading off buffer space (and consequently delay) and bandwidth. Our two-tiered bandwidth allocation scheme enables the delivery of any set of streams with less bandwidth (or equivalently with higher link utilization) than state-of-the-art deterministic smoothing approaches. The algorithm underlying our proposed frame-work uses three per-stream parameters and is linear in the number of servers, making it particularly well suited for use in an on-line setting. We present results from extensive trace-driven simulations, which confirm the efficiency of our scheme especially for small buffer sizes and delay bounds, and which underscore the significant realizable bandwidth savings, typically yielding losses that are an order of magnitude or more below our analytically derived bounds.

Indexing Methods for Efficient Multiclass Recognition

Many real world image analysis problems, such as face recognition and hand pose estimation, involve recognizing a large number of classes of objects or shapes. Large margin methods, such as AdaBoost and Support Vector Machines (SVMs), often provide competitive accuracy rates, but at the cost of evaluating a large number of binary classifiers, thus making it difficult to apply such methods when thousands or millions of classes need to be recognized. This thesis proposes a filter-and-refine framework, whereby, given a test pattern, a small number of candidate classes can be identified efficiently at the filter step, and computationally expensive large margin classifiers are used to evaluate these candidates at the refine step. Two different filtering methods are proposed, ClassMap and OVA-VS (One-vs.-All classification using Vector Search). ClassMap is an embedding-based method, works for both boosted classifiers and SVMs, and tends to map the patterns and their associated classes close to each other in a vector space. OVA-VS maps OVA classifiers and test patterns to vectors based on the weights and outputs of weak classifiers of the boosting scheme. At runtime, finding the strongest-responding OVA classifier becomes a classical vector search problem, where well-known methods can be used to gain efficiency. In our experiments, the proposed methods achieve significant speed-ups, in some cases up to two orders of magnitude, compared to exhaustive evaluation of all OVA classifiers. This was achieved in hand pose recognition and face recognition systems where the number of classes ranges from 535 to 48,600.

Indexing Distances in Large Graphs and Applications in Search Tasks

This thesis elaborates on the problem of preprocessing a large graph so that single-pair shortest-path queries can be answered quickly at runtime. Computing shortest paths is a well studied problem, but exact algorithms do not scale well to real-world huge graphs in applications that require very short response time. The focus is on approximate methods for distance estimation, in particular in landmarks-based distance indexing. This approach involves choosing some nodes as landmarks and computing (offline), for each node in the graph its embedding, i.e., the vector of its distances from all the landmarks. At runtime, when the distance between a pair of nodes is queried, it can be quickly estimated by combining the embeddings of the two nodes. Choosing optimal landmarks is shown to be hard and thus heuristic solutions are employed. Given a budget of memory for the index, which translates directly into a budget of landmarks, different landmark selection strategies can yield dramatically different results in terms of accuracy. A number of simple methods that scale well to large graphs are therefore developed and experimentally compared. The simplest methods choose central nodes of the graph, while the more elaborate ones select central nodes that are also far away from one another. The efficiency of the techniques presented in this thesis is tested experimentally using five different real world graphs with millions of edges; for a given accuracy, they require as much as 250 times less space than the current approach which considers selecting landmarks at random. Finally, they are applied in two important problems arising naturally in large-scale graphs, namely social search and community detection.

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.

Object Matching in Distributed Video Surveillance Systems by LDA-Based Appearance Descriptors

Establishing correspondences among object instances is still challenging in multi-camera surveillance systems, especially when the cameras’ fields of view are non-overlapping. Spatiotemporal constraints can help in solving the correspondence problem but still leave a wide margin of uncertainty. One way to reduce this uncertainty is to use appearance information about the moving objects in the site. In this paper we present the preliminary results of a new method that can capture salient appearance characteristics at each camera node in the network. A Latent Dirichlet Allocation (LDA) model is created and maintained at each node in the camera network. Each object is encoded in terms of the LDA bag-of-words model for appearance. The encoded appearance is then used to establish probable matching across cameras. Preliminary experiments are conducted on a dataset of 20 individuals and comparison against Madden’s I-MCHR is reported.

Optimal Scheduling of Secondary Content for Aggregation in Video-on-Demand Systems

Dynamic service aggregation techniques can exploit skewed access popularity patterns to reduce the costs of building interactive VoD systems. These schemes seek to cluster and merge users into single streams by bridging the temporal skew between them, thus improving server and network utilization. Rate adaptation and secondary content insertion are two such schemes. In this paper, we present and evaluate an optimal scheduling algorithm for inserting secondary content in this scenario. The algorithm runs in polynomial time, and is optimal with respect to the total bandwidth usage over the merging interval. We present constraints on content insertion which make the overall QoS of the delivered stream acceptable, and show how our algorithm can satisfy these constraints. We report simulation results which quantify the excellent gains due to content insertion. We discuss dynamic scenarios with user arrivals and interactions, and show that content insertion reduces the channel bandwidth requirement to almost half. We also discuss differentiated service techniques, such as N-VoD and premium no-advertisement service, and show how our algorithm can support these as well.

An Invariant Representation for Matching Trajectories across Uncalibrated Video Streams

We introduce a view-point invariant representation of moving object trajectories that can be used in video database applications. It is assumed that trajectories lie on a surface that can be locally approximated with a plane. Raw trajectory data is first locally approximated with a cubic spline via least squares fitting. For each sampled point of the obtained curve, a projective invariant feature is computed using a small number of points in its neighborhood. The resulting sequence of invariant features computed along the entire trajectory forms the view invariant descriptor of the trajectory itself. Time parametrization has been exploited to compute cross ratios without ambiguity due to point ordering. Similarity between descriptors of different trajectories is measured with a distance that takes into account the statistical properties of the cross ratio, and its symmetry with respect to the point at infinity. In experiments, an overall correct classification rate of about 95% has been obtained on a dataset of 58 trajectories of players in soccer video, and an overall correct classification rate of about 80% has been obtained on matching partial segments of trajectories collected from two overlapping views of outdoor scenes with moving people and cars.

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.

Learning Embeddings for Indexing, Retrieval, and Classification, with Applications to Object and Shape Recognition in Image Databases

Nearest neighbor retrieval is the task of identifying, given a database of objects and a query object, the objects in the database that are the most similar to the query. Retrieving nearest neighbors is a necessary component of many practical applications, in fields as diverse as computer vision, pattern recognition, multimedia databases, bioinformatics, and computer networks. At the same time, finding nearest neighbors accurately and efficiently can be challenging, especially when the database contains a large number of objects, and when the underlying distance measure is computationally expensive. This thesis proposes new methods for improving the efficiency and accuracy of nearest neighbor retrieval and classification in spaces with computationally expensive distance measures. The proposed methods are domain-independent, and can be applied in arbitrary spaces, including non-Euclidean and non-metric spaces. In this thesis particular emphasis is given to computer vision applications related to object and shape recognition, where expensive non-Euclidean distance measures are often needed to achieve high accuracy. The first contribution of this thesis is the BoostMap algorithm for embedding arbitrary spaces into a vector space with a computationally efficient distance measure. Using this approach, an approximate set of nearest neighbors can be retrieved efficiently - often orders of magnitude faster than retrieval using the exact distance measure in the original space. The BoostMap algorithm has two key distinguishing features with respect to existing embedding methods. First, embedding construction explicitly maximizes the amount of nearest neighbor information preserved by the embedding. Second, embedding construction is treated as a machine learning problem, in contrast to existing methods that are based on geometric considerations. The second contribution is a method for constructing query-sensitive distance measures for the purposes of nearest neighbor retrieval and classification. In high-dimensional spaces, query-sensitive distance measures allow for automatic selection of the dimensions that are the most informative for each specific query object. It is shown theoretically and experimentally that query-sensitivity increases the modeling power of embeddings, allowing embeddings to capture a larger amount of the nearest neighbor structure of the original space. The third contribution is a method for speeding up nearest neighbor classification by combining multiple embedding-based nearest neighbor classifiers in a cascade. In a cascade, computationally efficient classifiers are used to quickly classify easy cases, and classifiers that are more computationally expensive and also more accurate are only applied to objects that are harder to classify. An interesting property of the proposed cascade method is that, under certain conditions, classification time actually decreases as the size of the database increases, a behavior that is in stark contrast to the behavior of typical nearest neighbor classification systems. The proposed methods are evaluated experimentally in several different applications: hand shape recognition, off-line character recognition, online character recognition, and efficient retrieval of time series. In all datasets, the proposed methods lead to significant improvements in accuracy and efficiency compared to existing state-of-the-art methods. In some datasets, the general-purpose methods introduced in this thesis even outperform domain-specific methods that have been custom-designed for such datasets.

Extending snBench to Support a Video-Based Intrusion Detection and Alerting System with a Centralized Hash Table

In this project we design and implement a centralized hashing table in the snBench sensor network environment. We discuss the feasibility of this approach and compare and contrast with the distributed hashing architecture, with particular discussion regarding the conditions under which a centralized architecture makes sense. There are numerous computational tasks that require persistence of data in a sensor network environment. To help motivate the need for data storage in snBench we demonstrate a practical application of the technology whereby a video camera can monitor a room to detect the presence of a person and send an alert to the appropriate authorities.

ALD: adaptive layer distribution for scalable video

Recent years have witnessed a rapid growth in the demand for streaming video over the Internet and mobile networks, exposes challenges in coping with heterogeneous devices and varying network throughput. Adaptive schemes, such as scalable video coding, are an attractive solution but fare badly in the presence of packet losses. Techniques that use description-based streaming models, such as multiple description coding (MDC), are more suitable for lossy networks, and can mitigate the effects of packet loss by increasing the error resilience of the encoded stream, but with an increased transmission byte cost. In this paper, we present our adaptive scalable streaming technique adaptive layer distribution (ALD). ALD is a novel scalable media delivery technique that optimises the tradeoff between streaming bandwidth and error resiliency. ALD is based on the principle of layer distribution, in which the critical stream data are spread amongst all packets, thus lessening the impact on quality due to network losses. Additionally, ALD provides a parameterised mechanism for dynamic adaptation of the resiliency of the scalable video. The Subjective testing results illustrate that our techniques and models were able to provide levels of consistent high-quality viewing, with lower transmission cost, relative to MDC, irrespective of clip type. This highlights the benefits of selective packetisation in addition to intuitive encoding and transmission.

ALD: adaptive layer distribution for scalable video

Bandwidth constriction and datagram loss are prominent issues that affect the perceived quality of streaming video over lossy networks, such as wireless. The use of layered video coding seems attractive as a means to alleviate these issues, but its adoption has been held back in large part by the inherent priority assigned to the critical lower layers and the consequences for quality that result from their loss. The proposed use of forward error correction (FEC) as a solution only further burdens the bandwidth availability and can negate the perceived benefits of increased stream quality. In this paper, we propose Adaptive Layer Distribution (ALD) as a novel scalable media delivery technique that optimises the tradeoff between the streaming bandwidth and error resiliency. ALD is based on the principle of layer distribution, in which the critical stream data is spread amongst all datagrams thus lessening the impact on quality due to network losses. Additionally, ALD provides a parameterised mechanism for dynamic adaptation of the scalable video, while providing increased resilience to the highest quality layers. Our experimental results show that ALD improves the perceived quality and also reduces the bandwidth demand by up to 36% in comparison to the well-known Multiple Description Coding (MDC) technique.