1,184 research outputs found
CENSEgram-5M
No full textCreated By Félix Gontier and Mathieu Lagrange, LS2N, CNRS, Ecole Centrale Nantes
Contact : [email protected]
If used for research, please refer to:
@article{gontier2021training,
title={Polyphonic training set synthesis improves self-supervised urban sound classification},
author={Félix Gontier and Vincent Lostanlen, and Mathieu Lagrange and Nicolas Fortin and Jean-Francois Petiot and Catherine Lavandier},
journal={The Journal of the Acoustical Society of America},
year={2021},
publisher={Acoustical Society of America}
}
CENSEgram-5M contains third-octave spectrograms from the CENSE network of acoustic sensors.
These spectrograms correspond to five days of continuous measurements obtained in December 2019 by 16 sensors.
The total duration of the dataset is of the order of 5M seconds, i.e., 1280 hours.
The dataset is stored in a h5 file with one table called sensor that provides description of the sensors (id, latitude, longitude), and two groups that respectively store vectors of spectral and timestamp data.
The audio is made available as third octave spectral data, see demoTob.zip for an implementation of its computation from audio in Python.
From a python interpreter:
import tables as tb
>>> s=tb.open('CENSEgram-5M.h5')
Traceback (most recent call last):
File "", line 1, in
AttributeError: module 'tables' has no attribute 'open'
>>> s=tb.open_file('CENSEgram-5M.h5')
>>> print(s)
CENSEgram-5M.h5 (File) ''
Last modif.: 'Wed Apr 14 15:08:15 2021'
Object Tree:
/ (RootGroup) ''
/sensor (Table(16,)) 'Sensor information'
/spectrum (Group) 'spectral data third octave bands fast (125ms)'
/spectrum/sensor0 (Array(3296320, 29)) ''
...
/spectrum/sensor15 (Array(3296320, 29)) ''
...
/time (Group) 'time expressed in epoch'
/time/sensor0 (Array(3296320,)) ''
...
/time/sensor15 (Array(3296320,)) ''
The table sensor reads:
$ ptdump CENSEgram-5M.h5:/sensor -d
/sensor (Table(16,)) 'Sensor information'
Data dump:
[0] (b'urn:osh:sensor:noisemonitoring:B8-27-EB-E1-FB-4F', -3.36587, 47.74907, 0)
[1] (b'urn:osh:sensor:noisemonitoring:B8-27-EB-EA-EB-EA', -3.36526, 47.74832, 1)
[2] (b'urn:osh:sensor:noisemonitoring:B8-27-EB-A0-F5-4F', -3.36539, 47.74851, 2)
[3] (b'urn:osh:sensor:noisemonitoring:B8-27-EB-C3-D1-E5', -3.36452, 47.74914, 3)
[4] (b'urn:osh:sensor:noisemonitoring:B8-27-EB-77-1C-26', -3.36609, 47.74945, 4)
[5] (b'urn:osh:sensor:noisemonitoring:B8-27-EB-CC-42-7D', -3.36493, 47.74904, 5)
[6] (b'urn:osh:sensor:noisemonitoring:B8-27-EB-EA-12-88', -3.3684, 47.75114, 6)
[7] (b'urn:osh:sensor:noisemonitoring:B8-27-EB-5D-A9-60', -3.36774, 47.75078, 7)
[8] (b'urn:osh:sensor:noisemonitoring:B8-27-EB-60-92-E2', -3.36571, 47.74877, 8)
[9] (b'urn:osh:sensor:noisemonitoring:B8-27-EB-74-92-77', -3.36495, 47.7479, 9)
[10] (b'urn:osh:sensor:noisemonitoring:B8-27-EB-11-D0-A3', -3.35738, 47.75604, 10)
[11] (b'urn:osh:sensor:noisemonitoring:B8-27-EB-1F-AB-9F', -3.36477, 47.74782, 11)
[12] (b'urn:osh:sensor:noisemonitoring:B8-27-EB-4E-59-87', -3.36639, 47.74987, 12)
[13] (b'urn:osh:sensor:noisemonitoring:B8-27-EB-52-F4-03', -3.3651, 47.74806, 13)
[14] (b'urn:osh:sensor:noisemonitoring:B8-27-EB-20-DA-A9', -3.35946, 47.75488, 14)
[15] (b'urn:osh:sensor:noisemonitoring:B8-27-EB-56-87-4E', -3.3581, 47.75557, 15
Lorient-1k
No full text<p>Created By Félix Gontier and Mathieu Lagrange, LS2N, CNRS, Ecole Centrale Nantes</p>
<p>Contact : [email protected]</p>
<p>If used for research, please refer to:</p>
<pre>@article{gontier2021training,
title={Polyphonic training set synthesis improves self-supervised urban sound classification},
author={Félix Gontier and Vincent Lostanlen, and Mathieu Lagrange and Nicolas Fortin and Jean-Francois Petiot and Catherine Lavandier},
journal={The Journal of the Acoustical Society of America},
year={2021},
publisher={Acoustical Society of America}
}
</pre>
<p>Lorient-1k contains 30 acoustic scenes of duration equal to 45 seconds.<br>These scenes were recorded with Zoom H4n handheld devices at 10 different locations of Lorient (France).<br>Four experts annotated the onset and offset times of three sources of interest: traffic, voice, and birds. Those annotations have been taken into account to produce a single annotations that is coherent with the notion of perceived time of presence. That is, the sum of activations per scene and per source is coherent with the perceived time of presence.</p>
<p><br>The total duration of the dataset is of the order of 1.35k seconds, i.e., 22.5 minutes.</p>
<p>The audio is provided as third-octave spectral data and mel spectrograms (as of YAMNET). The audio is made available as third octave spectral data, see demoTob.zip for an implementation of its computation from audio in Python.</p>
<p> </p>
<p>From a python interpreter :</p>
<p>>> import numpy as np</p>
<p>>> s=np.load('Lorient-1k_spectralData.npy')</p>
<p>>> print(s.shape)</p>
<p>(30, 351, 29)</p>
<p>The three dimensions respectively corresponds to the sceneId, the frameId (time), and the spectralId (frequency).</p>
<p>>> a=np.load('Lorient-1k_presence.npy')</p>
<p>>> print(a.shape)</p>
<p>(30, 344, 3)</p>
<p>The third and fourth dimensions respectively corresponds to the sceneId, the frameId (time), the sourceId (traffic, voice, birds) and the annotatorId. Annotation is provided as a binary indicator of source presence for one second, that is 8 consecutive 125 ms frames with a hop of one frame.</p>
<p>>> a=np.load('Lorient-1k_time_of_presence.npy')</p>
<p>The time of presence is expressed in percents, per scene, and per source.</p>
<p>>> print(a.shape)</p>
<p>(30, 3)</p>
<p>The audio files are also available in the form of 16bits 44.1kHz wav files. Audio files are named in the same order as the first dimension of the .npy files : 00x.wav third-octaves and time of presence evaluation are accessed using s[x-1, :, : ] and a[x-1, :, : ]</p>
Lorient-1k
No full text<p>Created By Félix Gontier and Mathieu Lagrange, LS2N, CNRS, Ecole Centrale Nantes</p>
<p>Contact : [email protected]</p>
<p>If used for research, please refer to:</p>
<pre>@article{gontier2021training,
title={Polyphonic training set synthesis improves self-supervised urban sound classification},
author={Félix Gontier and Vincent Lostanlen, and Mathieu Lagrange and Nicolas Fortin and Jean-Francois Petiot and Catherine Lavandier},
journal={The Journal of the Acoustical Society of America},
year={2021},
publisher={Acoustical Society of America}
}
</pre>
<p>Lorient-1k contains 30 acoustic scenes of duration equal to 45 seconds.<br>These scenes were recorded with Zoom H4n handheld devices at 10 different locations of Lorient (France).<br>Four experts annotated the onset and offset times of three sources of interest: traffic, voice, and birds. Those annotations have been taken into account to produce a single annotations that is coherent with the notion of perceived time of presence. That is, the sum of activations per scene and per source is coherent with the perceived time of presence.</p>
<p><br>The total duration of the dataset is of the order of 1.35k seconds, i.e., 22.5 minutes.</p>
<p>The audio is provided as third-octave spectral data and mel spectrograms (as of YAMNET). The audio is made available as third octave spectral data, see demoTob.zip for an implementation of its computation from audio in Python.</p>
<p> </p>
<p>From a python interpreter :</p>
<p>>> import numpy as np</p>
<p>>> s=np.load('Lorient-1k_spectralData.npy')</p>
<p>>> print(s.shape)</p>
<p>(30, 351, 29)</p>
<p>The three dimensions respectively corresponds to the sceneId, the frameId (time), and the spectralId (frequency).</p>
<p>>> a=np.load('Lorient-1k_presence.npy')</p>
<p>>> print(a.shape)</p>
<p>(30, 344, 3)</p>
<p>The third and fourth dimensions respectively corresponds to the sceneId, the frameId (time), the sourceId (traffic, voice, birds) and the annotatorId. Annotation is provided as a binary indicator of source presence for one second, that is 8 consecutive 125 ms frames with a hop of one frame.</p>
<p>>> a=np.load('Lorient-1k_time_of_presence.npy')</p>
<p>The time of presence is expressed in percents, per scene, and per source.</p>
<p>>> print(a.shape)</p>
<p>(30, 3)</p>
<p>The audio files are also available in the form of 16bits 44.1kHz wav files. Audio files are named in the same order as the first dimension of the .npy files : 00x.wav third-octaves and time of presence evaluation are accessed using s[x-1, :, : ] and a[x-1, :, : ]</p>
Artificial-Delay Adaptive Control for Under-actuated Euler-Lagrange Robotics
No full textArtificial-delay control is a method in which state and input measurements collected at an immediate past time instant (i.e. artificially delayed) are used to compensate the uncertain dynamics affecting the system at the current time. This work formulates an artificial-delay control method with adaptive gains in the presence of nonlinear (Euler-Lagrange) under-actuation. The appeal of studying Euler-Lagrange dynamics is to capture many robotics applications of practical interest, as demonstrated via stability and robustness analysis and via robotic ship and robotic aerial vehicle test cases.Accepted Author ManuscriptTeam Bart De Schutte
Adaptive single-stage control for uncertain nonholonomic Euler-Lagrange systems
No full textThis work introduces a new single-stage adaptive controller for Euler-Lagrange systems with nonholonomic constraints. The proposed mechanism provides a simpler design philosophy compared to double-stage mechanisms (that address kinematics and dynamics in two steps), while achieving analogous stability properties, i.e. stability of both original and internal states. Meanwhile, we do not require direct access to the internal states as required in state-of-the-art single-stage mechanisms. The proposed approach is studied via Lyapunov analysis, validated numerically on wheeled mobile robot dynamics and compared to a standard double-stage approach.Green Open Access added to TU Delft Institutional Repository 'You share, we take care!' - Taverne project https://www.openaccess.nl/en/you-share-we-take-care Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public.Team Bart De Schutte
Dynamic constitutive equation of GFRP obtained by Lagrange experiment
No full textThe note presents a method of constructing dynamic constitutive equations of material by means of Lagrange experiment and analysis. Tests were carried out by a light gas gun and the stress history profiles were recorded on multiple Lagrange positions. The dynamic constitutive equations were deduced from the regression of a series of data which was obtained by Lagrange analysis based upon recorded multiple stress histories. Here constitutive equations of glass fibre reinforced phenolic resin composite(GFRP) in uniaxil strain state under dynamic loading are given. The proposed equations of the material agree well with experimental results
A comparative analysis of Lagrange multiplier and penalty approaches for modelling fluid-structure interaction
No full textPurpose: When simulating fluid-structure interaction (FSI), it is often essential that the no-slip condition is accurately enforced at the wetted boundary of the structure. This paper aims to evaluate the relative strengths and limitations of the penalty and Lagrange multiplier methods, within the context of modelling FSI, through a comparative analysis. Design/methodology/approach: In the immersed boundary method, the no-slip condition is typically imposed by augmenting the governing equations of the fluid with an artificial body force. The relative accuracy and computational time of the penalty and Lagrange multiplier formulations of this body force are evaluated by using each to solve three test problems, namely, flow through a channel, the harmonic motion of a cylinder through a stationary fluid and the vortex-induced vibration (VIV) of a cylinder. Findings: The Lagrange multiplier formulation provided an accurate solution, especially when enforcing the no-slip condition, and was robust as it did not require “tuning” of problem specific parameters. However, these benefits came at a higher computational cost relative to the penalty formulation. The penalty formulation achieved similar levels of accuracy to the Lagrange multiplier formulation, but only if the appropriate penalty factor was selected, which was difficult to determine a priori. Originality/value: Both the Lagrange multiplier and penalty formulations of the immersed boundary method are prominent in the literature. A systematic quantitative comparison of these two methods is presented within the same computational environment. A novel application of the Lagrange multiplier method to the modelling of VIV is also provided.Green Open Access added to TU Delft Institutional Repository ‘You share, we take care!’ – Taverne project https://www.openaccess.nl/en/you-share-we-take-care Otherwise as indicated in the copyright section: the publisher is the copyright holder of this work and the author uses the Dutch legislation to make this work public.Wind EnergyAerospace Structures & Computational Mechanic
A Spatial Bayesian Hedonic Pricing Model of Farmland Values
No full textIn 1973, British Columbia created the Agricultural Land Reserve (ALR) to protect farmland from development. This study investigates whether the ALR has been effective near the city of Victoria. Therefore, we employ a GIS-based hedonic pricing model and quantify ALR specific measures. Bayesian Model Averaging in combination with Markov Chain Monte Carlo Model Composition are used to address specification uncertainty. Results show that zoning schemes are partly credible. Zoned farmland sells for lower prices than other farmland. However, farmland located closer to the city of Victoria is priced higher and hobby farmers pay higher prices than conventional farmers.Farmland prices, Bayesian Model Averaging, Hedonic pricing., Land Economics/Use,
Lagrange-flux schemes: reformulating second-order accurate Lagrange-remap schemes for better node-based HPC performance
No full textInternational audienceIn a recent paper [Poncet R., Peybernes M., Gasc T., De Vuyst F. (2016) Performance modeling of a compressible hydrodynamics solver on multicore CPUs, in “Parallel Computing: on the road to Exascale”], we have achieved the performance analysis of staggered Lagrange-remap schemes, a class of solvers widely used for hydrodynamics applications. This paper is devoted to the rethinking and redesign of the Lagrange-remap process for achieving better performance using today’s computing architectures. As an unintended outcome, the analysis has lead us to the discovery of a new family of solvers – the so-called Lagrange-flux schemes – that appear to be promising for the CFD community.Dans un article récent [Poncet R., Peybernes M., Gasc T., De Vuyst F. (2016) Performance modeling of a compressible hydrodynamics solver on multicore CPUs, in “Parallel Computing: on the road to Exascale”], nous avons effectué l’analyse de la performance d’un schéma de type Lagrange+projection à variables décalées ; cette classe de solveurs est très utilisée pour les applications d’hydrodynamique. Dans cet article, on s’intéresse à la reformulation des solveurs Lagrange-projection afin d’améliorer leur performance globale sur architectures de calcul standards. De manière inattendue, l’analyse nous a conduit vers la découverte d’une nouvelle famille de solveurs – appelés schémas Lagrange-flux – qui apparaissent comme très prometteurs dans la communauté CFD
Lagrange-flux schemes: reformulating second-order accurate Lagrange-remap schemes for better node-based HPC performance
No full textIn a previous paper, we have achieved the performance analysis of staggered Lagrange-remap schemes, a class of solvers widely used for Hydrodynamics applications. This paper is devoted to the rethinking and redesign of the Lagrange-remap process for achieving better performance using today's computing architectures. As an unintended outcome, the analysis has lead us to the discovery of a new family of solvers -- the so-called Lagrange-flux schemes -- that appear to be promising for the CFD community
- …
