A low-power wireless video sensor node for distributed object detection

In this paper we propose MicrelEye, a wireless video node for cooperative distributed video processing applications that involve image classification. The node is equipped with a low-cost VGA CMOS image sensor, a reconfigurable processing engine (FPGA, Microcontroller, SRAM) and a Bluetooth 100-m transceiver. It has a size of few cubic centimeters and its typical power consumption is approximately ten times less than that of typical commercial DSP-based solutions. As regards classification, a highly optimized hardware-oriented support vector machine-like (SVM-like) algorithm called ERSVM is proposed and implemented. We describe our hardware and software architecture, its performance and power characteristics. The case study considered in this paper is people detection. The obtained results suggest that the present technology allows for the design of simple intelligent video nodes capable of performing classification tasks locally.

The evaporation signal from α particles stopped in superfluid helium

Alpha particles stopped in a 3 liter volume of liquid helium at 30 mK are observed by the calorimetric detection of helium atoms evaporated from the free surface of the liquid. Quantum evaporation of the helium is produced by the rotons that are created by the particle. While the energy spectrum of the 5.5 MeV 's from the²⁴¹Am source has a width of less than 0.5%, the energy distribution of the observed evaporation signals extends from the low energy threshold of several keV up to a maximum of several 100 keV, depending on geometrical factors and the collection area of the calorimeter. The origin of the observed distribution may result in part from the presence of the substrate and a dependence on the direction of the track of the particle. A simple model of the generation of rotons by the particle will be discussed.     

What hospitals expect today from clinical systems

Technology will always be a factor in hospital information system buys. But customers are increasingly looking for vendors that create a sincere, responsive sales relationship. In the end, the quality of support and understanding users receive can make or break a sale.     

Efficient additive kernels via explicit feature maps

Large scale nonlinear support vector machines (SVMs) can be approximated by linear ones using a suitable feature map. The linear SVMs are in general much faster to learn and evaluate (test) than the original nonlinear SVMs. This work introduces explicit feature maps for the additive class of kernels, such as the intersection, Hellinger's, and χ2 kernels, commonly used in computer vision, and enables their use in large scale problems. In particular, we: 1) provide explicit feature maps for all additive homogeneous kernels along with closed form expression for all common kernels; 2) derive corresponding approximate finite-dimensional feature maps based on a spectral analysis; and 3) quantify the error of the approximation, showing that the error is independent of the data dimension and decays exponentially fast with the approximation order for selected kernels such as χ2. We demonstrate that the approximations have indistinguishable performance from the full kernels yet greatly reduce the train/test times of SVMs. We also compare with two other approximation methods: Nystrom's approximation of Perronnin et al., which is data dependent, and the explicit map of Maji and Berg for the intersection kernel, which, as in the case of our approximations, is data independent. The approximations are evaluated on a number of standard data sets, including Caltech-101, Daimler-Chrysler pedestrians, and INRIA pedestrians.     

Multistage particle windows for fast and accurate object detection

The common paradigm employed for object detection is the sliding window (SW) search. This approach generates grid-distributed patches, at all possible positions and sizes, which are evaluated by a binary classifier: The tradeoff between computational burden and detection accuracy is the real critical point of sliding windows; several methods have been proposed to speed up the search such as adding complementary features. We propose a paradigm that differs from any previous approach since it casts object detection into a statistical-based search using a Monte Carlo sampling for estimating the likelihood density function with Gaussian kernels. The estimation relies on a multistage strategy where the proposal distribution is progressively refined by taking into account the feedback of the classifiers. The method can be easily plugged into a Bayesian-recursive framework to exploit the temporal coherency of the target objects in videos. Several tests on pedestrian and face detection, both on images and videos, with different types of classifiers (cascade of boosted classifiers, soft cascades, and SVM) and features (covariance matrices, Haar-like features, integral channel features, and histogram of oriented gradients) demonstrate that the proposed method provides higher detection rates and accuracy as well as a lower computational burden w.r.t. sliding window detection.     

Ontological query answering under expressive Entity-Relationship schemata

The Entity-Relationship (ER) model is a fundamental tool for database design, recently extended and employed in knowledge representation and reasoning due to its expressiveness and comprehensibility. We address the problem of answering conjunctive queries under constraints representing schemata expressed in an extended version of the Entity-Relationship model. This extended model, called ER⁺, comprises is-a constraints among entities and relationships, plus functional and mandatory participation constraints. In particular, it allows for arbitrary permutations of the roles in is-a among relationships. A key notion that ensures high tractability in ER⁺ schemata is separability, i.e., the absence of interaction between the functional participation constraints and the other constructs of ER⁺. We provide a precise syntactic characterization of separable ER⁺ schemata by means of a necessary and sufficient condition. We present a complete complexity analysis of the conjunctive query answering problem under separable ER⁺ schemata, and also under several sublanguages of ER⁺. We show that the addition of so-called negative constraints does not increase the complexity of query answering. With such constraints, our model properly generalizes the most widely adopted tractable ontology languages, including those in the DL-Lite family.     

Counting clicks and beeps: Exploring numerosity based haptic and audio PIN entry

Haptic and audio cues now appear commonly in computer interfaces, partially due to inherent advantages such as their support for eyes-free interaction. Their invisible, unobservable nature also makes them ideal candidates for security interfaces in which users have to enter secret information such as passwords. In particular, researchers have explored this idea through the design of PIN entry authentication systems based on multi-modal combinations of visual and non-visual content or on the recognition of small sets of unimodal haptic or audio stimuli. This paper highlights the benefits and performance limitations of these approaches and introduces an alternative based on unimodal audio or haptic temporal numerosity – the ability to accurately and rapidly determine the number of cues presented in rapid temporal succession. In essence, in a numerosity interface, rather than recognizing distinct cues, users must count the number of times that a single cue occurs. In an iterative process of design and evaluation, three prototypes implementing this concept are presented and studies of their use reported. The results show the fastest PIN entry times and lowest error rates to be 8 s and 2%, figures that improve substantially on previous research. These results are attained while maintaining low levels of workload and substantial resistance to observation attack (as determined via camera attack security studies). In sum, this paper argues that unimodal audio and haptic numerosity is a valuable and relatively unexplored metaphor for non-visual input and demonstrates the validity of this claim in the demanding task of unobservable authentication systems.     

Self-Organising Maps: A new way to screen the level of satisfaction of dialysis patients

Evaluation of patient satisfaction has become an important indicator for assessing health care quality. Fresenius Medical Care (FME) as a global provider of dialysis services through its NephroCare network has a strong interest in monitoring patient satisfaction.The aim of the paper is to test and validate a methodology for detecting a residual area of low satisfaction in dialysis patients.The FME Patient Satisfaction Programme questionnaire was distributed to haemodialysis (HD) patients treated in 335 centers of its network. It contained 79 questions covering various satisfaction aspects regarding Dialysis Unit, Dialysis Arrangement, Nurses, Doctors, etc.To analyse the data provided by the questionnaire, the Self-Organising Map (SOM) method was used. SOM is a neural network model for clustering and projecting high-dimensional data into a low-dimensional space, preserving topological relationships of original high-dimensional data spaces.10,632 HD patients completed the questionnaire. Mean age was 63.05 ± 14.93 years with 56.69% males. Response rate was 66%. Overall level of satisfaction was 1.99 (range from ?3 to +3). On average patients were very satisfied with all issues. Nevertheless, a group of patients, around 60 years old, balanced gender ratio, whose level of satisfaction was lower than 1, were highlighted.In the NephroCare clinics patient satisfaction with service is rather high. While traditional analysis usually stops here, the SOM method allows identification of areas of potential improvement for specific patient groups.     

Discrete time sliding mode control of robotic manipulators: Development and experimental validation

This paper presents a robust discrete-time sliding mode control coupled with an uncertainty estimator designed for planar robotic manipulators. Experimental evidence shows satisfactory trajectory tracking performances and noticeable robustness in the presence of model inaccuracies, disturbances and payload perturbations. Ultimate boundedness of the tracking errors is proved, as well as boundedness of the estimation error with arbitrary precision.     

Ultimate behavior of axially loaded RC wall-like columns confined with GFRP

The assessment of the effectiveness of the fiber reinforced polymer (FRP) confinement on rectangular reinforced concrete (RC) columns with high aspect ratio (wall-like) still represents an unresolved issue. The present paper aims at providing more experimental evidence about the behavior of such members confined with both uni-directional and quadri-directional glass FRP laminates. Particular attention is devoted to issues related to the premature failure of confining fibers experimentally observed in wall-like columns. Test results on nine axially loaded columns are herein presented; emphasis is also given to the analysis of FRP strain profiles along the sides of the cross-section. The analysis of test results highlights that glass FRP (GFRP) confinement could determine significant strength and ductility increases; the discussion of failure modes points out that the failure of GFRP confined wall-like columns is controlled by the shape of the cross-section and occurs at transverse strains in the jacket much lower than those ultimate of the fibers. Theoretical–experimental comparisons are performed using some available models for strength prediction of such members.