Search

The recent emergence of new applications in the area of wireless video sensor network and ultra-low-power biomedical applications (such as the wireless camera pill) have created new design challenges and frontiers requiring extensive research work. In such applications, it is often required to capture a large amount of data and process them in real time while the hardware is constrained to take very little physical space and to consume very little power. This is only possible using custom single-chip solutions integrating image sensor and hardware-friendly image compression algorithms. This paper proposes an adaptive quantization scheme based on boundary adaptation procedure followed by an online quadrant tree decomposition processing enabling low power and yet robust and compact image compression processor integrated together with a digital CMOS image sensor. The image sensor chip has been implemented using 0.35-[micro]m CMOS technology and operates at 3.3 V. Simulation and experimental results show compression figures corresponding to 0.6-0.8 bit per pixel, while maintaining reasonable peak signal-to-noise ratio levels and very low operating power consumption. In addition, the proposed compression processor is expected to benefit significantly from higher resolution and Megapixels CMOS imaging technology.

Ehrlichia ruminantium is the causative agent of heartwater, a major tick-borne disease of livestock in Africa that has been introduced in the Caribbean and is threatening to emerge and spread on the American mainland. We sequenced the complete genomes of two strains of E. ruminantium of differing phenotypes, strains Gardel (Erga; 1,499,920 bp), from the island of Guadeloupe, and Welgevonden (Erwe; 1,512,977 bp), originating in South Africa and maintained in Guadeloupe in a different cell environment. Comparative genomic analysis of these two strains was performed with the recently published parent strain of Erwe (Erwo) and other Rickettsiales (Anaplasma, Wolbachia, and Rickettsia spp.). Gene order is highly conserved between the E. ruminantium strains and with A. marginale. In contrast, there is very little conservation of gene order with members of the Rickettsiaceae. However, gene order may be locally conserved, as illustrated by the tuf operons. Eighteen truncated protein-encoding sequences (CDSs) differentiate Erga from Erwe/Erwo, whereas four other truncated CDSs differentiate Erwe from Erwo. Moreover, E. ruminantium displays the lowest coding ratio observed among bacteria due to unusually long intergenic regions. This is related to an active process of genome expansion/contraction targeted at tandem repeats in noncoding regions and based on the addition or removal of ca. 150-bp tandem units. This process seems to be specific to E. ruminantium and is not observed in the other Rickettsiales.

Ehrlichia ruminantium, an obligate intracellular bacterium transmitted by ticks of the genus Amblyomma, causes heartwater disease in ruminants. The gene coding for the major antigenic protein MAP1 is part of a multigene family consisting of a cluster containing 16 paralogs. In the search for differentially regulated genes between E. ruminantium grown in endothelial and tick cell lines that could be used in vaccine development and to determine if differences in the map1 gene cluster exist between different isolates of E. ruminantium, we analyzed the map1 gene cluster of the Senegal and Gardel isolates of E. ruminantium. Both isolates contained the same number of genes, and the same organization as found in the genome sequence of the Welgevonden isolate (H. Van Heerden, N. E. Collins, K. A. Brayton, C. Rademeyer, and B. A. Allsopp, Gene 330:159-168, 2004). However, comparison of two subpopulations of the Gardel isolate maintained in different laboratories demonstrated that recombination between map1-3 and map1-2 had occurred in one subpopulation with deletion of one entire gene. Reverse transcription-PCR on E. ruminantium derived mRNA from infected cells using gene-specific primers revealed that all 16 map1 paralogs were transcribed in endothelial cells. In one vector (Amblyomma variegatum) and several nonvector tick cell lines infected with E. ruminantium, transcripts were found for between 4 and 11 paralogs. In all these cases the transcript for the map1-1 gene was detected and was predominant. Our results indicate that the map1 gene cluster is relatively conserved but can be subject to recombination, and differences in the transcription of map1 multigenes in host and vector cell environments exist.

We conducted extensive surveillance for West Nile virus infection in equines and chickens in Guadeloupe in 2003-2004. We showed a high seroprevalence in equines in 2003 related to biome, followed [...]

A bagging ensemble consists of a set of classifiers trained independently and combined by a majority vote. Such a combination improves generalization performance but can require large amounts of memory and computation, a serious drawback for addressing portable real-time pattern recognition applications. We report here a compact three-dimensional (3-D) multiprecision very large-scale integration (VLSI) implementation of a bagging ensemble. In our circuit, individual classifiers are decision trees implemented as threshold networks-one layer of threshold logic units (TLUs) followed by combinatorial logic functions. The hardware was fabricated using 0.7-[micro]m CMOS technology and packaged using MCM-V micro-packaging technology. The 3-D chip implements up to 192 TLUs operating at a speed of up to 48GCPPS and implemented in a volume of (w x L x h) = (2 x 2 x 0.7) [cm.sup.3]. The 3-D circuit features a high level of programmability and flexibility offering the possibility to make an efficient use of the hardware resources in order to reduce the power consumption. Successful operation of the 3-D chip for various precisions and ensemble sizes is demonstrated through an electronic nose application. Index Terms--Bagging, decision trees, threshold networks, very large-scale integration (VLSI), three-dimensional (3-D) packaging technology.

We derive a new method for solving nonlinear blind source separation (BSS) problems by exploiting second-order statistics in a kernel induced feature space. This paper extends a new and efficient closed-form linear algorithm to the nonlinear domain using the kernel trick originally applied in support vector machines (SVMs). This technique could likewise be applied to other linear covariance-based source separation algorithms. Experiments on realistic nonlinear mixtures of speech signals, gas multisensor data, and visual disparity data illustrate the applicability of our approach. Index Terms--Blind source separation (BSS), support vector machines (SVMs).

Traditional analog-to-digital (A/D) conversion is a limited scalar coding technique, unfit for compression, but which can be implemented with very fast, simple, and small analog devices. This brief extends the concept of A/D conversion from scalars to vectors. A vector A/D converter offers a level of complexity similar to that of usual scalar A/D converters but operates on a block of k analog inputs so as to perform a mixed conversion/compression task. To adapt to changing source statistics, a backward, unsupervised learning rule is proposed. The rule, called [BAR.sub.r](k). attempts to minimize the rth power law distortion in the high resolution case. Adaptation is only determined by the transmitted codeword and can be performed simultaneously at both sides of the channel without any side information. The learning rule includes a forgetting factor to increase robustness in case of transmission errors and initial encoder/decoder mismatches. Index Terms - Backward adaptation, data structures, digital-analog conversion, tree data structures, unsupervised learning, vector quantization.

In this paper, a neural competitive learning tree is introduced as a computationally attractive scheme for adaptive density estimation and novelty detection. The learning rule yields equiprobable quantizations of the input space and provides an adaptive focusing mechanism capable of tracking time-varying distributions. It is shown by simulation that the neural tree performs reasonably well while being much faster than any of the other competitive learning algorithms. Index Terms - Adaptive multivariable histogram, density estimation, equiprobable quantization, sequential detection, tree data structures, unsupervised learning.

This letter presents a novel unsupervised competitive learning rule called the boundary adaptation rule (BAR), for scalar quantization. It is shown both mathematically and by simulations that BAR converges to equiprobable quantizations of univariate probability density functions and that, in this way, it outperforms other unsupervised competitive learning rules.

To determine whether West Nile virus (WNV) had reached the archipelago of Guadeloupe, a serologic study in horses and birds was conducted in 2002. immunoglobulin (Ig) G, IgM, enzyme-linked immunosorbent [...]

Author(s): Dominique Martinez [1] Like many other insects, a female moth releases specific blends of odours -- pheromones -- to signal her presence to males. The pheromone plume is not [...]