Search

International audience; Humans use whistled communications, the most elaborate of which are commonly called “whistled languages” or “whistled speech” because they consist of a natural type of speech. The principle of whistled speech is straightforward: people articulate words while whistling and thereby transform spoken utterances by simplifying them, syllable by syllable, into whistled melodies. One of the most striking aspects of this whistled transformation of words is that it remains intelligible to trained speakers, despite a reduced acoustic channel to convey meaning. It constitutes a natural traditional means of telecommunication that permits spoken communication at long distances in a large diversity of languages of the world. Historically, birdsong has been used as a model for vocal learning and language. But conversely, human whistled languages can serve as a model for elucidating how information may be encoded in dolphin whistle communication. In this paper, we elucidate the reasons why human whistled speech and dolphin whistles are interesting to compare. Both are characterized by similar acoustic parameters and serve a common purpose of long distance communication in natural surroundings in two large brained social species. Moreover, their differences – e.g., how they are produced, the dynamics of the whistles, and the types of information they convey – are not barriers to such a comparison. On the contrary, by exploring the structure and attributes found across human whistle languages, we highlight that they can provide an important model as to how complex information is and can be encoded in what appears at first sight to be simple whistled modulated signals. Observing details, such as processes of segmentation and coarticulation, in whistled speech can serve to advance and inform the development of new approaches for the analysis of whistle repertoires of dolphins, and eventually other species. Human whistled languages and dolphin whistles could serve as complementary test benches for the development of new methodologies and algorithms for decoding whistled communication signals by providing new perspectives on how information may be encoded structurally and organizationally.

Tracking the origin of propagating wave signals in an environment with complex reflective surfaces is, in its full generality, a nearly intractable problem which has engendered multiple domain-specific literatures. We posit that, if the environment and sensor geometries are fixed, machine learning algorithms can "learn" the acoustical geometry of the environment and accurately track signal origin. In this paper, we propose the first machine-learning-based approach to identifying the source locations of semi-stationary, tonal, dolphin-whistle-like sounds in a highly reverberant space, specifically a half-cylindrical dolphin pool. Our algorithm works by supplying a learning network with an overabundance of location "clues", which are then selected under supervised training for their ability to discriminate source location in this particular environment. More specifically, we deliver estimated time-difference-of-arrivals (TDOA's) and normalized cross-correlation values computed from pairs of hydrophone signals to a random forest model for high-feature-volume classification and feature selection, and subsequently deliver the selected features into linear discriminant analysis, linear and quadratic Support Vector Machine (SVM), and Gaussian process models. Based on data from 14 sound source locations and 16 hydrophones, our classification models yielded perfect accuracy at predicting novel sound source locations. Our regression models yielded better accuracy than the established Steered-Response Power (SRP) method when all training data were used, and comparable accuracy along the pool surface when deprived of training data at testing sites; our methods additionally boast improved computation time and the potential for superior localization accuracy in all dimensions with more training data. Because of the generality of our method we argue it may be useful in a much wider variety of contexts.

In daily life, in the operating room and in the laboratory, the operational way to assess wakefulness and consciousness is through responsiveness. A number of studies suggest that the awake, conscious state is not the default behavior of an assembly of neurons, but rather a very special state of activity that has to be actively maintained and curated to support its functional properties. Thus responsiveness is a feature that requires active maintenance, such as a homeostatic mechanism to balance excitation and inhibition. In this work we developed a method for monitoring such maintenance processes, focusing on a specific signature of their behavior derived from the theory of dynamical systems: stability analysis of dynamical modes. When such mechanisms are at work, their modes of activity are at marginal stability, neither damped (stable) nor exponentially growing (unstable) but rather hovering in between. We have previously shown that, conversely, under induction of anesthesia those modes become more stable and thus less responsive, then reversed upon emergence to wakefulness. We take advantage of this effect to build a single-trial classifier which detects whether a subject is awake or unconscious achieving high performance. We show that our approach can be developed into a means for intra-operative monitoring of the depth of anesthesia, an application of fundamental importance to modern clinical practice.

In recent years, considerable evidence has accrued indicating that brain function can be flexibly reconfigured on the fly: many brain areas are capable of carrying out a variety of distinct functions and are able to switch between those functions in a context-sensitive, dynamic fashion. Some evidence has also emerged that ongoing brain activity, the ceaseless background brain dynamics, may be implicated in setting and controlling these flexible functions as well as attentional gain. A crucial link between them is dynamics. Here we review both the accumulating evidence as well as propose a theoretical outline of dynamical mechanisms for functional self-reconfiguration of neural networks, including reconfiguration of logic function, reconfiguration of information routing, and poising at critical points to switch dynamics.

Electrophysiology has recently evolved into an interactive, high-throughput endeavor. Recording from dozens to hundreds of electrodes is today routine; novel means of manipulating the system in real time, through electrical stimulation, optogenetics or sensory manipulation are allowing us to decipher neural circuit function at an unparalleled rate. To contribute to the wide dissemination of such techniques, we present an open hardware project, High-Channel Count Electrophysiology (HiCCE), aiming to produce low-cost, high-channel count (≥128 channels) electrophysiology instrumentation capable of fast data acquisition rates, real-time processing and feedback capabilities. Our design is centered on an open standard, FPGA Mezzanine Card (FMC), which permits a varied choice of FPGA carrier architectures suited to different laboratory experimental needs. The HiCCE-128, a low-cost highperformance 128-channel data acquisition board for small voltage signals, is being introduced. It is a FMC module that can be operated from any FPGA carrier conforming to the FMC/VITA57 standard. This specialized board with a low input referred noise of about 3 μV is capable of acquiring 128 channels simultaneously at 31.25 kS/s per channel with 16 effective bits of resolution. We present the global architecture and some preliminary measurement to illustrate its potential for electrophysiological and medical applications.

The brain processes visual inputs having structure over a large range of spatial scales. The precise mechanisms or algorithms used by the brain to achieve this feat are largely unknown and an open problem in visual neuroscience. In particular, the spatial extent in visual space over which primary visual cortex (V1) performs evidence integration has been shown to change as a function of contrast and other visual parameters, thus adapting scale in visual space in an input-dependent manner. We demonstrate that a simple dynamical mechanism---dynamical criticality---can simultaneously account for the well-documented input-dependence characteristics of three properties of V1: scales of integration in visuotopic space, extents of lateral integration on the cortical surface, and response latencies.

We recently presented an estimate of the number of mutually discriminable olfactory stimuli at one trillion (1). Subjects were asked to sniff mixtures of molecules with increasing component overlap selected from a panel of 128 isointense structurally and perceptually diverse monomolecular odorants (2). We considered stimulus pairs discriminable when the majority of subjects could significantly discriminate them at p=0.05, a conventional statistical threshold given our sample size. From these empirical data, we estimated that human discriminative capacity exceeds one trillion olfactory stimuli. Several readers have pointed out that such extrapolations are sensitive to underlying assumptions about the chosen significance threshold (3) and the dimensionality of olfaction (4). It is important to note that any exponential function will be sensitive in this way, and the goal of our model was not to identify the exact number of discriminable olfactory stimuli, or even the exact mathematical bounds, but an estimate of the order of magnitude of human discriminatory power across a population of human subjects. This was not clearly stated in our paper, and we agree that contradictory references to a “lower limit” and an “upper bound” were confusing. The central argument in (4) is that our estimation method assumes that the dimensionality of olfactory space is large. We agree that the high-dimensional nature of olfaction is indeed an assumption, and we should have stated this explicitly in our paper (1). Even if we follow this logic of the models presented in (4), purely geometrical calculations show that our results hold if the dimensionality of olfactory representations is D≥25. The dimensionality of olfaction is a question of interest to everyone, and while we do not know for sure, all available evidence suggests that olfaction is a high-dimensional sense. The olfactory system is wired to keep information from the ~400 odorant receptors strictly separated, so it is plausible that olfaction operates at least in 400-dimensional space. This is an important topic of discussion in olfaction, and we welcome continued debate of the dimensionality of smell and how this impacts human olfactory perception.

We construct a LIF-type stochastic model for interspike times of the firing activity of a single neuron subject to a time-varying input signal. By using first passage time densities some numerical evaluations of ISI densities and comparisons with simulation results are given.

The firing activity of intracellularly stimulated neurons in cortical slices has been demonstrated to be profoundly affected by the temporal structure of the injected current (Mainen & Sejnowski, 1995 ). This suggests that the timing features of the neural response may be controlled as much by its own biophysical characteristics as by how a neuron is wired within a circuit. Modeling studies have shown that the interplay between internal noise and the fluctuations of the driving input controls the reliability and the precision of neuronal spiking (Cecchi et al., 2000 ; Tiesinga, 2002 ; Fellous, Rudolph, Destexhe, & Sejnowski, 2003 ). In order to investigate this interplay, we focus on the stochastic leaky integrate-and-fire neuron and identify the Hölder exponent H of the integrated input as the key mathematical property dictating the regime of firing of a single-unit neuron. We have recently provided numerical evidence (Taillefumier & Magnasco, 2013 ) for the existence of a phase transition when [Formula: see text] becomes less than the statistical Hölder exponent associated with internal gaussian white noise (H=1/2). Here we describe the theoretical and numerical framework devised for the study of a neuron that is periodically driven by frozen noisy inputs with exponent H>0. In doing so, we account for the existence of a transition between two regimes of firing when H=1/2, and we show that spiking times have a continuous density when the Hölder exponent satisfies H>1/2. The transition at H=1/2 formally separates rate codes, for which the neural firing probability varies smoothly, from temporal codes, for which the neuron fires at sharply defined times regardless of the intensity of internal noise.

To understand how the human visual system analyzes images, it is essential to know the structure of the visual environment. In particular, natural images display consistent statistical properties that distinguish them from random luminance distributions. We have studied the geometric regularities of oriented elements (edges or line segments) present in an ensemble of visual scenes, asking how much information the presence of a segment in a particular location of the visual scene carries about the presence of a second segment at different relative positions and orientations. We observed strong long-range correlations in the distribution of oriented segments that extend over the whole visual field. We further show that a very simple geometric rule, cocircularity, predicts the arrangement of segments in natural scenes, and that different geometrical arrangements show relevant differences in their scaling properties. Our results show similarities to geometric features of previous physiological and psychophysical studies. We discuss the implications of these findings for theories of early vision., Comment: 3 figures, 2 large figures not included

We show that presentation of individual canary song syllables results in distinct expression patterns of the immediate-early gene ZENK in the caudomedial neostriatum (NCM) of adult canaries. Information on the spatial distribution and labeling of stained cells provides for a classification of ZENK patterns that (1) accords to the organization of stimuli into families, (2) preserves the stimuli intrafamily relationships, and (3) confers salience to natural over artificial stimuli, resulting in a nonclassical tonotopic map. Moreover, complex syllable maps cannot be reduced to any linear combinations of simple syllable maps. These properties arise from the collective response of NCM neurons to auditory stimuli, rather than from the behavior of single neurons. The syllabic representation described here may constitute an important step toward deciphering the rules of birdsong auditory representation.

Crystallization of two-dimensional domains in monolayer films of the soluble surfactant sodium dodecyl sulfate exhibits a smooth shape transition from disks to hexagons. This transition is experimentally proven to be mainly a kinetic effect. The appearance of negative curvature marks the onset of shape instabilities. A simple geometric model based on the local interplay between line tension and crystalline anisotropy accounts for the observed behaviors. it is argued that the observed shapes are generic and insensitive to details of the actual mechanisms

Microscopists are familiar with many blemishes that fluorescence images can have due to dust and debris, glass flaws, uneven distribution of fluids or surface coatings, etc. Microarray scans do show similar artifacts, which might affect subsequent analysis. We developed a tool, Harshlight, for the detection and masking of blemishes in HDONA microarray chips. Harshlight uses a combination of statistic and image processing methods to identify defects. We demonstrate that Harshlight can be widely used for chips with different technologies thanks to its user-tunable parameters. Here we report its application to SNP microarrays.

Auditory neurons preserve exquisite temporal information about sound features, but we do not know how the brain uses this information to process the rapidly changing sounds of the natural world. Simple arguments for effective use of temporal information led us to consider the reassignment class of time-frequency representations as a model of auditory processing. Reassigned time-frequency representations can track isolated simple signals with accuracy unlimited by the time-frequency uncertainty principle, but lack of a general theory has hampered their application to complex sounds. We describe the reassigned representations for white noise and show that even spectrally dense signals produce sparse reassignments: the representation collapses onto a thin set of lines arranged in a froth-like pattern. Preserving phase information allows reconstruction of the original signal. We define a notion of “consensus,” based on stability of reassignment to time-scale changes, which produces sharp spectral estimates for a wide class of complex mixed signals. As the only currently known class of time-frequency representations that is always “in focus” this methodology has general utility in signal analysis. It may also help explain the remarkable acuity of auditory perception. Many details of complex sounds that are virtually undetectable in standard sonograms are readily perceptible and visible in reassignment.

The tuning curve of the cochlea measures how intense an input is required to elicit a given output level as a function of the frequency. It is a fundamental object of auditory theory, for it summarizes how to identify sounds on the basis of the cochlear output. A simple model is presented showing that only two elements are sufficient for establishing the cochlear tuning curve: a broadly tuned traveling wave, moving unidirectionally from high to low frequencies, and a set of mechanosensors poised at the threshold of an oscillatory (Hopf) instability. These two components generate the various frequency-response regimes needed for a cochlear tuning curve with a high slope.

We present a model of sound production in a songbird’s vocal organ and find that much of the complexity of the song of the canary (Serinus canaria) can be produced from simple time variations in forcing functions. The starts, stops, and pauses between syllables, as well as variation in pitch and timbre are inherent in the mechanics and can often be expressed through smooth and simple variations in the frequency and relative phase of two driving parameters Human language and the song of many bird species are both learned by juveniles through experience. The ontogeny of learned song begins with early food-begging calls, continues through more complex bablings known as subsong, to eventually reach an adult form of song which is first plastic, and later stereotyped [1]. This development depends on auditory experience, for abnormal forms of song develop if a juvenile is deafened and unable to hear its own sound production or is isolated from the experience of any adult song model to copy. The time course of development and the dependence on auditory experience are similar to aspects of human speech, suggesting there may be common principles of learning and memory under

Background Microscopists are familiar with many blemishes that fluorescence images can have due to dust and debris, glass flaws, uneven distribution of fluids or surface coatings, etc. Microarray scans do show similar artifacts, which might affect subsequent analysis. Although all but the starkest blemishes are hard to find by the unaided eye, particularly in high-density oligonucleotide arrays (HDONAs), few tools are available to help with the detection of those defects. Results We develop a novel tool, Harshlight, for the automatic detection and masking of blemishes in HDONA microarray chips. Harshlight uses a combination of statistic and image processing methods to identify three different types of defects: localized blemishes affecting a few probes, diffuse defects affecting larger areas, and extended defects which may invalidate an entire chip. Conclusion We demonstrate the use of Harshlight can materially improve analysis of HDONA chips, especially for experiments with subtle changes between samples. For the widely used MAS5 algorithm, we show that compact blemishes cause an average of 8 gene expression values per chip to change by more than 50%, two of them by more than twofold; our masking algorithm restores about two thirds of this damage. Large-scale artifacts are successfully detected and eliminated.