Commissariat &#xc0; L'&#xc9;nergie Atomique Et Aux &#xc9;nergies Alternatives

Ulis

Centre National De La Recherche Scientifique

Université Paris-sud 11

Université Paris 13

Université Paris—sud

Antoine Dupret

Arnaud Verdant

William Guicquero

Esteban Cabanillas

Michel Vilain

Olivier Bichler

Benoît Dupont

Vincent Lorrain

Timothé Laforest

Aurélie Touvignon

Suzanne Lesecq

Diego Puschini Pascual

Sylvain Gigan

Laurent Alacoque

Gilles Tessier

Nicolas Ravot

Christophe Layer

Wafa Ben Hassen

Julien Mottin

Assim Boukhayma

Pierre Vincent

Jean-Michel Tualle

Jaume Benoit

Miguel Gallego Roman

François Ramaz

Tien Ho Phuoc

Marius Vasiliu

Armando Zanchetta

Benoit Dupont

Hugues Duvillier

Aurelie Touvignon

A method for generating a reflectometry signal intended to be injected into a transmission line in order to identify the presence of at least one possible fault on the line, the method includes the steps of: generating a first multi-carrier digital signal on a first set of frequency carriers, generating a second corrective multi-carrier digital signal on a second set of frequency carriers separate from the first set, summing the first digital signal and the second corrective digital signal so as to generate a reflectometry signal, normalizing the frequency carriers of the reflectometry signal, the second corrective digital signal being determined so as to reduce the peak to average power ratio of the reflectometry signal with respect to the first digital signal.

Method for reducing the peak power of a multi-carrier reflectometry signal

A processor for computing at least one convolution layer of a convolutional neural network is provided, in response to an input event, the convolutional neural network comprising at least one convolution kernel, the convolution kernel containing weight coefficients. The processor comprises at least one convolution module configured to compute the one or more convolution layers, each convolution module comprising a set of elementary processing units for computing the internal value of the convolution-layer neurons that are triggered by the input event, each convolution module being configured to match the weight coefficients of the kernel with certain at least of the elementary processing units of the module in parallel, the number of elementary processing units being independent of the number of neurons of the convolution layer.

Device and method for calculating convolution in a convolutional neural network

A device for distributing the convolution coefficients of the least one convolutional kernel of a convolutional neural network is provided, the coefficients being carried by an input bus, to a set of processing units in a processor based on a convolutional-neural-network architecture. The device comprises at least one switching network that is controlled by at least one control unit, the switching network comprising a set of switches that are arranged to apply circular shifts to at least one portion of the input bus. For each convolution kernel, each control unit is configured to dynamically control certain at least of the switches of the switching networks in response to an input event applied to the convolution kernel and at least one parameter representing the maximum size of the convolution kernels.

Device and method for distributing convolutional data of a convolutional neural network

A system for analyzing faults in a transmission line includes a test signal generator, a coupler for injecting the test signal into the transmission line and sampling the signal back-propagated in the transmission line, a control unit for generating a calibration signal equal to the opposite of the test signal and controlling its phase and its amplitude, at least one combiner for combining the calibration signal and the signal sampled by the coupler into a corrected signal, an amplifier for amplifying the corrected signal and a correlator for correlating the corrected signal and the test signal to produce a reflectogram, the control unit being further configured to make the amplitude and the phase of the calibration signal vary so as to eliminate, in the signal sampled by the coupler, at least one echo of the test signal on at least one point of impedance mismatch of the transmission line.

System for analysing faults by reflectometry of optimised dynamic range

A method and device for detecting faults in a transmission line by reflectometry, include the following steps: injecting into the transmission line a reference signal at an emission frequency f; collecting a reflected signal at a point on the transmission line; sampling the reflected signal at a sampling frequency f, the sampling frequency fbeing different from the emission frequency f; storing each point of the sampled signal at a memory address corresponding to an index assigned to the point of the sampled signal and according to a precomputed memory-address increment &#x394;, the memory-address increment &#x394; depending on the emission frequency f, on the sampling frequency f, on an over-sampling factor &#x3a9; and on a preset acquisition time &#x3a3;; repeating the storing step during the acquisition time &#x3a3;; and generating, from the points stored during the acquisition time, a recomposed signal able to be used to detect faults.

Method and device for detecting faults in a transmission line

A method for analyzing faults affecting a transmission line network includes at least one line segment having a plurality of ends, the method comprising the steps of: acquiring, at at least two ends of the line network, a measurement of a signal being propagated in the network, applying a time reversal to each signal measurement, for each segment of the line network, constructing a cost function dependent on the time and the distance from a point of the segment to an end of the network, searching for at least one extremum of the cost function over all of the segments of the line network and retaining the coordinates of the extremum, deducing therefrom the presence of a fault at a position and an instant of occurrence that are given by the coordinates of the extremum.

Method and device for deducing presence of faults in a transmission line network

A method for detecting soft faults in a transmission line includes the following steps: acquiring a measurement, called time-domain reflectogram, of a signal characteristic of the reflection of a reference signal previously injected into the line, determining the difference between the time-domain reflectogram and a time-domain reflectogram measured previously for the same line or an identical line of similar characteristics, in order to obtain a corrected time-domain reflectogram, applying a plurality of independent transformations to the corrected time-domain reflectogram in order to obtain a plurality of independent transformed reflectograms, converting the transformed reflectograms into a plurality of mutually independent probabilities of occurrence of a fault, applying a data merging method to the probabilities of occurrence of a fault to deduce therefrom a unified value of the probability of occurrence of a fault.

Method for detecting soft faults in a cable by data fusion

A system for acquiring images by means of heterodyne digital holography comprises an image sensor having at least one photodiode coupled to an oversampling analog-digital converter.

System for acquiring images by means of heterodyne digital holography

The present invention involves a CMOS image sensor including: a plurality of pixels distributed in a plurality of subsets, each subset defining a measurement support; and a circuit capable of implementing a compressive image sensing method, the method including: for a given integration time of the sensor pixels, reading, for each measurement support, a single output value based on the output values of a pseudo-random selection of one or a plurality of pixels of the support.

Compressive image sensing device and method

A system for acquiring a hyperspectral image, including: a grey level image sensor; and a diffuser and dispersive element placed on the optical path between the sensor and a scene, this element including an array of individually-controllable liquid crystal cells, where each cell can receive a control voltage selected from among a series of at least three different control voltages.

Hyperspectral image acquisition system and method

The invention concerns a system for measuring and controlling the wave-front of a coherent light beam (), comprising: a device for generating a reference light beam () that is coherent with said coherent light beam (); a partially reflective monolithic device () comprising an array of elementary cells, each elementary cell comprising at least one element () for measuring a portion of the incident wave interfering between the coherent light beam and the reference light beam and a phase modulation element () for modulating the phase of the reflected beam; and a control device () for controlling said phase modulation element making use of the information from the associated measuring element.

Device for measuring and controlling the wave-front of a coherent light beam

A method of acquiring an image of a scene illuminated by a first beam, by means of a sensor including at least two pixels, including the steps of: a) for each pixel, reading a first output value of the pixel representative of the intensity of the radiation received by the pixel during a first integration period during which the sensor is illuminated by a second beam coherent with the first beam; and b) for each pixel, reading a second output value of the pixel representative of the intensity of the radiation received by the pixel during a second integration period during which the sensor is not illuminated by the second beam, wherein steps a) and b) are repeated a plurality of times by changing, between two successive iterations, the angle of incidence or a phase parameter of the second beam.

Image acquisition method and system

A CMOS image sensor including: a plurality of pixels; a first analog circuit for calculating one or a plurality of statistical estimators based on the analog output values of sensor pixels; and a second circuit capable of implementing a compressive image sensing method, wherein the applied compression rate is a function of the statistical estimator(s) calculated by the first circuit.

The invention relates to a method for generating super-resolution images using a processing device, the method includes: oversampling an input image to generate an oversampled image; modifying the pixel value of a first pixel of the oversampled image by: generating a similarity value for each one of a plurality of candidate pixels in the input image and/or in one or more other input images, candidate pixels being selected based on the position of the first pixel in the oversampled image, similarity value being generated by evaluating the similarity of a group of pixels adjacent to each of the candidate pixels to a group of pixels adjacent to the first pixel; selecting a first subset of candidate pixels on the basis of similarity values, and generating the modified pixel value of the first pixel based on the pixel values and on the similarity values of the first subset of candidate pixels.

Method and device for reconstructing super-resolution images

This analog electronic circuit () for processing a light signal (), of the type comprising:

Growing community of inventors

Orsay, France

Antoine Dupret