Scene Classification Using a Hybrid Generative/Discriminative Approach

We investigate whether dimensionality reduction using a latent generative model is beneficial for the task of weakly supervised scene classification. In detail, we are given a set of labeled images of scenes (for example, coast, forest, city, river, etc.), and our objective is to classify a new image into one of these categories. Our approach consists of first discovering latent ";topics"; using probabilistic Latent Semantic Analysis (pLSA), a generative model from the statistical text literature here applied to a bag of visual words representation for each image, and subsequently, training a multiway classifier on the topic distribution vector for each image. We compare this approach to that of representing each image by a bag of visual words vector directly and training a multiway classifier on these vectors. To this end, we introduce a novel vocabulary using dense color SIFT descriptors and then investigate the classification performance under changes in the size of the visual vocabulary, the number of latent topics learned, and the type of discriminative classifier used (k-nearest neighbor or SVM). We achieve superior classification performance to recent publications that have used a bag of visual word representation, in all cases, using the authors' own data sets and testing protocols. We also investigate the gain in adding spatial information. We show applications to image retrieval with relevance feedback and to scene classification in videos

An integrated genetic and functional analysis of the role of type II transmembrane serine proteases (TMPRSSs) in hearing loss.

Building on our discovery that mutations in the transmembrane serine protease, TMPRSS3, cause nonsyndromic deafness, we have investigated the contribution of other TMPRSS family members to the auditory function. To identify which of the 16 known TMPRSS genes had a strong likelihood of involvement in hearing function, three types of biological evidence were examined: 1) expression in inner ear tissues; 2) location in a genomic interval that contains a yet unidentified gene for deafness; and 3) evaluation of hearing status of any available Tmprss knockout mouse strains. This analysis demonstrated that, besides TMPRSS3, another TMPRSS gene was essential for hearing and, indeed, mice deficient for Hepsin (Hpn) also known as Tmprss1 exhibited profound hearing loss. In addition, TMPRSS2, TMPRSS5, and CORIN, also named TMPRSS10, showed strong likelihood of involvement based on their inner ear expression and mapping position within deafness loci PKSR7, DFNB24, and DFNB25, respectively. These four TMPRSS genes were then screened for mutations in affected members of the DFNB24 and DFNB25 deafness families, and in a cohort of 362 sporadic deaf cases. This large mutation screen revealed numerous novel sequence variations including three potential pathogenic mutations in the TMPRSS5 gene. The mutant forms of TMPRSS5 showed reduced or absent proteolytic activity. Subsequently, TMPRSS genes with evidence of involvement in deafness were further characterized, and their sites of expression were determined. Tmprss1, 3, and 5 proteins were detected in spiral ganglion neurons. Tmprss3 was also present in the organ of Corti. TMPRSS1 and 3 proteins appeared stably anchored to the endoplasmic reticulum membranes, whereas TMPRSS5 was also detected at the plasma membrane. Collectively, these results provide evidence that TMPRSS1 and TMPRSS3 play and TMPRSS5 may play important and specific roles in hearing.

An Instrumental analysis of the European Union´s capability to act in conflict response

Since its origins, the European Union has striven to be an actor on the International scene and a place in conflict Management. Yet the EU’s lack of activity cannot be justified by a mere lack of capacities. The EU counts with numerous political, economic, and, since 2003, civil and military instruments that should allow it to precede a comprehensive conflict response. This publication consists of a description of these instruments and an analysis of the final use that the Union makes of them in the different stages of a conflict. Examples will show us the EU’s main weakness in providing a comprehensive and timely response when a conflict breaks out.

Auditory-visual multisensory interactions in humans: timing, topography, directionality, and sources.

Current models of brain organization include multisensory interactions at early processing stages and within low-level, including primary, cortices. Embracing this model with regard to auditory-visual (AV) interactions in humans remains problematic. Controversy surrounds the application of an additive model to the analysis of event-related potentials (ERPs), and conventional ERP analysis methods have yielded discordant latencies of effects and permitted limited neurophysiologic interpretability. While hemodynamic imaging and transcranial magnetic stimulation studies provide general support for the above model, the precise timing, superadditive/subadditive directionality, topographic stability, and sources remain unresolved. We recorded ERPs in humans to attended, but task-irrelevant stimuli that did not require an overt motor response, thereby circumventing paradigmatic caveats. We applied novel ERP signal analysis methods to provide details concerning the likely bases of AV interactions. First, nonlinear interactions occur at 60-95 ms after stimulus and are the consequence of topographic, rather than pure strength, modulations in the ERP. AV stimuli engage distinct configurations of intracranial generators, rather than simply modulating the amplitude of unisensory responses. Second, source estimations (and statistical analyses thereof) identified primary visual, primary auditory, and posterior superior temporal regions as mediating these effects. Finally, scalar values of current densities in all of these regions exhibited functionally coupled, subadditive nonlinear effects, a pattern increasingly consistent with the mounting evidence in nonhuman primates. In these ways, we demonstrate how neurophysiologic bases of multisensory interactions can be noninvasively identified in humans, allowing for a synthesis across imaging methods on the one hand and species on the other.

An analysis of visual adaptation and contrast perception for tone mapping

Tone Mapping is the problem of compressing the range of a High-Dynamic Range image so that it can be displayed in a Low-Dynamic Range screen, without losing or introducing novel details: The final image should produce in the observer a sensation as close as possible to the perception produced by the real-world scene. We propose a tone mapping operator with two stages. The first stage is a global method that implements visual adaptation, based on experiments on human perception, in particular we point out the importance of cone saturation. The second stage performs local contrast enhancement, based on a variational model inspired by color vision phenomenology. We evaluate this method with a metric validated by psychophysical experiments and, in terms of this metric, our method compares very well with the state of the art.

Generalized linear and generalized additive models in studies of species distributions: setting the scene

An important statistical development of the last 30 years has been the advance in regression analysis provided by generalized linear models (GLMs) and generalized additive models (GAMs). Here we introduce a series of papers prepared within the framework of an international workshop entitled: Advances in GLMs/GAMs modeling: from species distribution to environmental management, held in Riederalp, Switzerland, 6-11 August 2001.We first discuss some general uses of statistical models in ecology, as well as provide a short review of several key examples of the use of GLMs and GAMs in ecological modeling efforts. We next present an overview of GLMs and GAMs, and discuss some of their related statistics used for predictor selection, model diagnostics, and evaluation. Included is a discussion of several new approaches applicable to GLMs and GAMs, such as ridge regression, an alternative to stepwise selection of predictors, and methods for the identification of interactions by a combined use of regression trees and several other approaches. We close with an overview of the papers and how we feel they advance our understanding of their application to ecological modeling.

Auditory motion affects visual biological motion processing.

The processing of biological motion is a critical, everyday task performed with remarkable efficiency by human sensory systems. Interest in this ability has focused to a large extent on biological motion processing in the visual modality (see, for example, Cutting, J. E., Moore, C., & Morrison, R. (1988). Masking the motions of human gait. Perception and Psychophysics, 44(4), 339-347). In naturalistic settings, however, it is often the case that biological motion is defined by input to more than one sensory modality. For this reason, here in a series of experiments we investigate behavioural correlates of multisensory, in particular audiovisual, integration in the processing of biological motion cues. More specifically, using a new psychophysical paradigm we investigate the effect of suprathreshold auditory motion on perceptions of visually defined biological motion. Unlike data from previous studies investigating audiovisual integration in linear motion processing [Meyer, G. F. & Wuerger, S. M. (2001). Cross-modal integration of auditory and visual motion signals. Neuroreport, 12(11), 2557-2560; Wuerger, S. M., Hofbauer, M., & Meyer, G. F. (2003). The integration of auditory and motion signals at threshold. Perception and Psychophysics, 65(8), 1188-1196; Alais, D. & Burr, D. (2004). No direction-specific bimodal facilitation for audiovisual motion detection. Cognitive Brain Research, 19, 185-194], we report the existence of direction-selective effects: relative to control (stationary) auditory conditions, auditory motion in the same direction as the visually defined biological motion target increased its detectability, whereas auditory motion in the opposite direction had the inverse effect. Our data suggest these effects do not arise through general shifts in visuo-spatial attention, but instead are a consequence of motion-sensitive, direction-tuned integration mechanisms that are, if not unique to biological visual motion, at least not common to all types of visual motion. Based on these data and evidence from neurophysiological and neuroimaging studies we discuss the neural mechanisms likely to underlie this effect.

GPS analysis of human locomotion: further evidence for long-range correlations in stride-to-stride fluctuations of gait parameters.

During free walking, gait is automatically adjusted to provide optimal mechanical output and minimal energy expenditure; gait parameters, such as cadence, fluctuate from one stride to the next around average values. It was described that this fluctuation exhibited long-range correlations and fractal-like patterns. In addition, it was suggested that these long-range correlations disappeared if the participant followed the beep of metronome to regulate his or her pace. Until now, these fractal fluctuations were only observed for stride interval, because no technique existed to adequately analyze an extended time of free walking. The aim of the present study was to measure walking speed (WS), step frequency (SF) and step length (SL) with high accuracy (<1 cm) satellite positioning method (global positioning system or GPS) in order to detect long-range correlations in the stride-to-stride fluctuations. Eight participants walked 30 min under free and constrained (metronome) conditions. Under free walking conditions, DFA (detrended fluctuation analysis) and surrogate data tests showed that the fluctuation of WS, SL and SF exhibited a fractal pattern (i.e., scaling exponent alpha: 0.5 < alpha < 1) in a large majority of participants (7/8). Under constrained conditions (metronome), SF fluctuations became significantly anti-correlated (alpha < 0.5) in all participants. However, the scaling exponent of SL and WS was not modified. We conclude that, when the walking pace is controlled by an auditory signal, the feedback loop between the planned movement (at supraspinal level) and the sensory inputs induces a continual shifting of SF around the mean (persistent anti-correlation), but with no effect on the fluctuation dynamics of the other parameters (SL, WS).

Single-trial topographic analysis of human EEG: A new image of event-related potentials

We present a novel approach for analyzing single-trial electroencephalography (EEG) data, using topographic information. The method allows for visualizing event-related potentials using all the electrodes of recordings overcoming the problem of previous approaches that required electrode selection and waveforms filtering. We apply this method to EEG data from an auditory object recognition experiment that we have previously analyzed at an ERP level. Temporally structured periods were statistically identified wherein a given topography predominated without any prior information about the temporal behavior. In addition to providing novel methods for EEG analysis, the data indicate that ERPs are reliably observable at a single-trial level when examined topographically.

Tonotopic mapping of human auditory cortex.

Since the early days of functional magnetic resonance imaging (fMRI), retinotopic mapping emerged as a powerful and widely-accepted tool, allowing the identification of individual visual cortical fields and furthering the study of visual processing. In contrast, tonotopic mapping in auditory cortex proved more challenging primarily because of the smaller size of auditory cortical fields. The spatial resolution capabilities of fMRI have since advanced, and recent reports from our labs and several others demonstrate the reliability of tonotopic mapping in human auditory cortex. Here we review the wide range of stimulus procedures and analysis methods that have been used to successfully map tonotopy in human auditory cortex. We point out that recent studies provide a remarkably consistent view of human tonotopic organisation, although the interpretation of the maps continues to vary. In particular, there remains controversy over the exact orientation of the primary gradients with respect to Heschl's gyrus, which leads to different predictions about the location of human A1, R, and surrounding fields. We discuss the development of this debate and argue that literature is converging towards an interpretation that core fields A1 and R fold across the rostral and caudal banks of Heschl's gyrus, with tonotopic gradients laid out in a distinctive V-shaped manner. This suggests an organisation that is largely homologous with non-human primates. This article is part of a Special Issue entitled Human Auditory Neuroimaging.

The role of actions in auditory object discrimination.

Action representations can interact with object recognition processes. For example, so-called mirror neurons respond both when performing an action and when seeing or hearing such actions. Investigations of auditory object processing have largely focused on categorical discrimination, which begins within the initial 100 ms post-stimulus onset and subsequently engages distinct cortical networks. Whether action representations themselves contribute to auditory object recognition and the precise kinds of actions recruiting the auditory-visual mirror neuron system remain poorly understood. We applied electrical neuroimaging analyses to auditory evoked potentials (AEPs) in response to sounds of man-made objects that were further subdivided between sounds conveying a socio-functional context and typically cuing a responsive action by the listener (e.g. a ringing telephone) and those that are not linked to such a context and do not typically elicit responsive actions (e.g. notes on a piano). This distinction was validated psychophysically by a separate cohort of listeners. Beginning approximately 300 ms, responses to such context-related sounds significantly differed from context-free sounds both in the strength and topography of the electric field. This latency is >200 ms subsequent to general categorical discrimination. Additionally, such topographic differences indicate that sounds of different action sub-types engage distinct configurations of intracranial generators. Statistical analysis of source estimations identified differential activity within premotor and inferior (pre)frontal regions (Brodmann's areas (BA) 6, BA8, and BA45/46/47) in response to sounds of actions typically cuing a responsive action. We discuss our results in terms of a spatio-temporal model of auditory object processing and the interplay between semantic and action representations.

Separating mismatch negativity (MMN) response from auditory obligatory brain responses in school-aged children.

Mismatch negativity (MMN) overlaps with other auditory event-related potential (ERP) components. We examined the ERPs of 50 9- to 11-year-old children for vowels /i/, /y/ and equivalent complex tones. The goal was to separate MMN from obligatory ERP components using principal component analysis and equal probability control condition. In addition to the contrast of the deviant minus standard response, we employed the contrast of the deviant minus control response, to see whether the obligatory processing contributes to MMN in children. When looking for differences in speech deviant minus standard contrast, MMN starts around 112 ms. However, when both contrasts are examined, MMN emerges for speech at 160 ms whereas for nonspeech MMN is observed at 112 ms regardless of contrast. We argue that this discriminative response to speech stimuli at 112 ms is obligatory in nature rather than reflecting change detection processing.

Intelligence-led crime scene processing. Part I : Forensic intelligence.

Forensic science is generally defined as the application of science to address questions related to the law. Too often, this view restricts the contribution of science to one single process which eventually aims at bringing individuals to court while minimising risk of miscarriage of justice. In order to go beyond this paradigm, we propose to refocus the attention towards traces themselves, as remnants of a criminal activity, and their information content. We postulate that traces contribute effectively to a wide variety of other informational processes that support decision making inmany situations. In particular, they inform actors of new policing strategies who place the treatment of information and intelligence at the centre of their systems. This contribution of forensic science to these security oriented models is still not well identified and captured. In order to create the best condition for the development of forensic intelligence, we suggest a framework that connects forensic science to intelligence-led policing (part I). Crime scene attendance and processing can be envisaged within this view. This approach gives indications abouthowto structure knowledge used by crime scene examiners in their effective practice (part II).

Radarsat backscattering from an agricultural scene

Orbital remote sensing in the microwave electromagnetic region has been presented as an important tool for agriculture monitoring. The satellite systems in operation have almost all-weather capability and high spatial resolution, which are features appropriated for agriculture. However, for full exploration of these data, an understanding of the relationships between the characteristics of each system and agricultural targets is necessary. This paper describes the behavior of backscattering coefficient (sigma°) derived from calibrated data of Radarsat images from an agricultural area. It is shown that in a dispersion diagram of sigma° there are three main regions in which most of the fields can be classified. The first one is characterized by low backscattering values, with pastures and bare soils; the second one has intermediate backscattering coefficients and comprises well grown crops mainly; and a third one, with high backscattering coefficients, in which there are fields with strong structures causing a kind of double bounce effect. The results of this research indicate that the use of Radarsat images is optimized when a multitemporal analysis is done making the best use of the agricultural calendar and of the dynamics of different cultures.

Pre-stimulus beta oscillations within left posterior sylvian regions impact auditory temporal order judgment accuracy.

Both neural and behavioral responses to stimuli are influenced by the state of the brain immediately preceding their presentation, notably by pre-stimulus oscillatory activity. Using frequency analysis of high-density electroencephalogram coupled with source estimations, the present study investigated the role of pre-stimulus oscillatory activity in auditory spatial temporal order judgments (TOJ). Oscillations within the beta range (i.e. 18-23Hz) were significantly stronger before accurate than inaccurate TOJ trials. Distributed source estimations identified bilateral posterior sylvian regions as the principal contributors to pre-stimulus beta oscillations. Activity within the left posterior sylvian region was significantly stronger before accurate than inaccurate TOJ trials. We discuss our results in terms of a modulation of sensory gating mechanisms mediated by beta activity.