Estimating 3D Hand Pose from a Cluttered Image

A method is proposed that can generate a ranked list of plausible three-dimensional hand configurations that best match an input image. Hand pose estimation is formulated as an image database indexing problem, where the closest matches for an input hand image are retrieved from a large database of synthetic hand images. In contrast to previous approaches, the system can function in the presence of clutter, thanks to two novel clutter-tolerant indexing methods. First, a computationally efficient approximation of the image-to-model chamfer distance is obtained by embedding binary edge images into a high-dimensional Euclide an space. Second, a general-purpose, probabilistic line matching method identifies those line segment correspondences between model and input images that are the least likely to have occurred by chance. The performance of this clutter-tolerant approach is demonstrated in quantitative experiments with hundreds of real hand images.

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.

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.

VAMP - a vision based sensor network for health care hygiene

Adequate hand-washing has been shown to be a critical activity in preventing the transmission of infections such as MRSA in health-care environments. Hand-washing guidelines published by various health-care related institutions recommend a technique incorporating six hand-washing poses that ensure all areas of the hands are thoroughly cleaned. In this paper, an embedded wireless vision system (VAMP) capable of accurately monitoring hand-washing quality is presented. The VAMP system hardware consists of a low resolution CMOS image sensor and FPGA processor which are integrated with a microcontroller and ZigBee standard wireless transceiver to create a wireless sensor network (WSN) based vision system that can be retargeted at a variety of health care applications. The device captures and processes images locally in real-time, determines if hand-washing procedures have been correctly undertaken and then passes the resulting high-level data over a low-bandwidth wireless link. The paper outlines the hardware and software mechanisms of the VAMP system and illustrates that it offers an easy to integrate sensor solution to adequately monitor and improve hand hygiene quality. Future work to develop a miniaturized, low cost system capable of being integrated into everyday products is also discussed.

Estimating stochastic volatility diffusion using conditional moments of integrated volatility

We exploit the distributional information contained in high-frequency intraday data in constructing a simple conditional moment estimator for stochastic volatility diffusions. The estimator is based on the analytical solutions of the first two conditional moments for the latent integrated volatility, the realization of which is effectively approximated by the sum of the squared high-frequency increments of the process. Our simulation evidence indicates that the resulting GMM estimator is highly reliable and accurate. Our empirical implementation based on high-frequency five-minute foreign exchange returns suggests the presence of multiple latent stochastic volatility factors and possible jumps. © 2002 Elsevier Science B.V. All rights reserved.

Compressive holography.

Compressive sampling enables signal reconstruction using less than one measurement per reconstructed signal value. Compressive measurement is particularly useful in generating multidimensional images from lower dimensional data. We demonstrate single frame 3D tomography from 2D holographic data.

Seasonal ITCZ migration dynamically controls the location of the (sub)tropical Atlantic biogeochemical divide

Inorganic nitrogen depletion restricts productivity in much of the low-latitude oceans, generating a selective advantage for diazotrophic organisms capable of fixing atmospheric dinitrogen (N2). However, the abundance and activity of diazotrophs can in turn be controlled by the availability of other potentially limiting nutrients, including phosphorus (P) and iron (Fe). Here we present high-resolution data (∼0.3°) for dissolved iron, aluminum, and inorganic phosphorus that confirm the existence of a sharp north–south biogeochemical boundary in the surface nutrient concentrations of the (sub)tropical Atlantic Ocean. Combining satellite-based precipitation data with results from a previous study, we here demonstrate that wet deposition in the region of the intertropical convergence zone acts as the major dissolved iron source to surface waters. Moreover, corresponding observations of N2 fixation and the distribution of diazotrophic Trichodesmium spp. indicate that movement in the region of elevated dissolved iron as a result of the seasonal migration of the intertropical convergence zone drives a shift in the latitudinal distribution of diazotrophy and corresponding dissolved inorganic phosphorus depletion. These conclusions are consistent with the results of an idealized numerical model of the system. The boundary between the distinct biogeochemical systems of the (sub)tropical Atlantic thus appears to be defined by the diazotrophic response to spatial–temporal variability in external Fe inputs. Consequently, in addition to demonstrating a unique seasonal cycle forced by atmospheric nutrient inputs, we suggest that the underlying biogeochemical mechanisms would likely characterize the response of oligotrophic systems to altered environmental forcing over longer timescales.

Greenberger-Horne-Zeilinger nonlocality in arbitrary even dimensions

We generalize Greenberger-Horne-Zeilinger (GHZ) nonlocality to every even-dimensional and odd-partite system. For the purpose we employ concurrent observables that are incompatible and nevertheless have a common eigenstate. It is remarkable that a tripartite system can exhibit the genuinely high-dimensional GHZ nonlocality.

Predicting Cell Cycle Regulated Genes by Causal Interactions

The fundamental difference between classic and modern biology is that technological innovations allow to generate high-throughput data to get insights into molecular interactions on a genomic scale. These high-throughput data can be used to infer gene networks, e. g., the transcriptional regulatory or signaling network, representing a blue print of the current dynamical state of the cellular system. However, gene networks do not provide direct answers to biological questions, instead, they need to be analyzed to reveal functional information of molecular working mechanisms. In this paper we propose a new approach to analyze the transcriptional regulatory network of yeast to predict cell cycle regulated genes. The novelty of our approach is that, in contrast to all other approaches aiming to predict cell cycle regulated genes, we do not use time series data but base our analysis on the prior information of causal interactions among genes. The major purpose of the present paper is to predict cell cycle regulated genes in S. cerevisiae. Our analysis is based on the transcriptional regulatory network, representing causal interactions between genes, and a list of known periodic genes. No further data are used. Our approach utilizes the causal membership of genes and the hierarchical organization of the transcriptional regulatory network leading to two groups of periodic genes with a well defined direction of information flow. We predict genes as periodic if they appear on unique shortest paths connecting two periodic genes from different hierarchy levels. Our results demonstrate that a classical problem as the prediction of cell cycle regulated genes can be seen in a new light if the concept of a causal membership of a gene is applied consequently. This also shows that there is a wealth of information buried in the transcriptional regulatory network whose unraveling may require more elaborate concepts than it might seem at first.

Commonality in Returns, Order Flows, and Liquidity in the Greek Stock Market

Using a unique high-frequency data-set on a comprehensive sample of Greek blue-chip stocks, spanning from September 2003 through March 2006, this note assesses the extent and role of commonality in returns, order flows (OFs), and liquidity. It also formally models aggregate equity returns in terms of aggregate equity OF, in an effort to clarify OF's importance in explaining returns for the Athens Exchange market. Almost a quarter of the daily returns in the FTSE/ATHEX20 index is explained by aggregate own OF. In a second step, using principal components and canonical correlation analyses, we document substantial common movements in returns, OFs, and liquidity, both on a market-wide basis and on an individual security basis. These results emphasize that asset pricing and liquidity cannot be analyzed in isolation from each other.

Behavioral plasticity allows short-term adjustment to a novel environment

Many species are currently experiencing anthropogenically driven environmental changes. Among these changes, increasing noise levels are specifically a problem for species relying on acoustic communication. Recent evidence suggests that some species adjust their acoustic signals to man-made noise. However, it is unknown whether these changes occur through short-term and reversible adjustments by behavioral plasticity or through long-term adaptations by evolutionary change. Using behavioral observations and playback experiments, we show that male reed buntings (Emberiza schoeniclus) adjusted their songs immediately, singing at a higher minimum frequency and at a lower rate when noise levels were high. Our data showed that these changes in singing behavior were short-term adjustments of signal characteristics resulting from behavioral plasticity, rather than a long-term adaptation. However, more males remained unpaired at a noisy location than at a quiet location throughout the breeding season. Thus, phenotypic plasticity enables individuals to respond to environmental changes, but whether these short-term adjustments are beneficial remains to be seen.

Inferring the conservative causal core of gene regulatory networks

Background
Inferring gene regulatory networks from large-scale expression data is an important problem that received much attention in recent years. These networks have the potential to gain insights into causal molecular interactions of biological processes. Hence, from a methodological point of view, reliable estimation methods based on observational data are needed to approach this problem practically.

Results
In this paper, we introduce a novel gene regulatory network inference (GRNI) algorithm, called C3NET. We compare C3NET with four well known methods, ARACNE, CLR, MRNET and RN, conducting in-depth numerical ensemble simulations and demonstrate also for biological expression data from E. coli that C3NET performs consistently better than the best known GRNI methods in the literature. In addition, it has also a low computational complexity. Since C3NET is based on estimates of mutual information values in conjunction with a maximization step, our numerical investigations demonstrate that our inference algorithm exploits causal structural information in the data efficiently.

Conclusions
For systems biology to succeed in the long run, it is of crucial importance to establish methods that extract large-scale gene networks from high-throughput data that reflect the underlying causal interactions among genes or gene products. Our method can contribute to this endeavor by demonstrating that an inference algorithm with a neat design permits not only a more intuitive and possibly biological interpretation of its working mechanism but can also result in superior results.

Structural influence of gene networks on their inference: analysis of C3NET

Background: The availability of large-scale high-throughput data possesses considerable challenges toward their functional analysis. For this reason gene network inference methods gained considerable interest. However, our current knowledge, especially about the influence of the structure of a gene network on its inference, is limited.

Detection of Voigt Spectral Line Profiles of Hydrogen Radio Recombination Lines toward Sagittarius B2(N)

We report the detection of Voigt spectral line profiles of radio recombination lines (RRLs) toward Sagittarius B2(N) with the 100 m Green Bank Telescope (GBT). At radio wavelengths, astronomical spectra are highly populated with RRLs, which serve as ideal probes of the physical conditions in molecular cloud complexes. An analysis of the Hn alpha lines presented herein shows that RRLs of higher principal quantum number (n > 90) are generally divergent from their expected Gaussian profiles and, moreover, are well described by their respective Voigt profiles. This is in agreement with the theory that spectral lines experience pressure broadening as a result of electron collisions at lower radio frequencies. Given the inherent technical difficulties regarding the detection and profiling of true RRL wing spans and shapes, it is crucial that the observing instrumentation produce flat baselines as well as high-sensitivity, high-resolution data. The GBT has demonstrated its capabilities regarding all of these aspects, and we believe that future observations of RRL emission via the GBT will be crucial toward advancing our knowledge of the larger-scale extended structures of ionized gas in the interstellar medium (ISM).