Vous êtes ici : Accueil » Kiosque » Annonce


Mot de passe : 

Mot de passe oublié ?
Détails d'identification oubliés ?


16 juillet 2018

Analysis of cyclostationary signals based on deep learning methodologies

Catégorie : Doctorant

Starting date: November or December 2018

Application deadline date: September 5th 2018

Decision announcement date: September 20th 2018

In the ADEME IMOTEP project context (Innovation MOTEur Propre https://www.univ-st-etienne.fr/fr/tous-les-faits-marquants/annees-precedentes/annee-2016-2017/zoom-sur/imotep.html - see at the bottom of this web page for a presentation of the project partners of this project), this PhD thesis subject is proposed by the Image team of the Hubert Curien laboratory(https://laboratoirehubertcurien.univ-st-etienne.fr/en/teams/image-science-computer-vision.html) in collaboration with the “Laboratory of Tribologie and Dynamic of Systems” (LTDS - http://ltds.ec-lyon.fr/spip/) of the “Ecole Centrale” in Lyon and IREIS (Institut de Recherche En Ingénierie des Surfaces), the research and development department of HEF group (https://www.hef.fr/).

Cyclostationary signals are random signals whose statistical characteristics have periodicities [GN06]. It is possible to encounter signals of this type in many fields such as telecommunications, radar signals, acoustic signals, mechanical signals related to rotations in engines, etc. Note that it is also possible to model with this kind of approaches random patterns that may have spatial periodicities : the cyclostationary signals are not limited to signals indexed by the time as soon as the nature of the support is different (spatial, spatio-temporal, …).

Whatever the field, the problems addressed by the scientific community concerning these signals are often related to the denoising and the separation of a mixture of several signals (or sources). One can notice that the denoising task can also be seen as the separation of the useful signal from the noise [BA97, AMX01]. It is also possible to consider improving the considered methods by detecting (or recognizing) the types of present signals, which then makes it possible to exploit the most appropriate models.

Recently, new methods of denoising and source separation, exploiting the current advances in machine learning based on deep learning (LSTM-RNN [KY17], deep NN [KK18], Autoencoders [GP17], ...) have emerged for speech or audio signal processing. Our main objective is to extend this work to cyclostationary signals and apply the developed methods to friction measurement signals (tribological measurements) acquired on test benches for the engines of the future.

The PhD work will start with the realization of a state of the art, on the one hand, on the methods of denoising and source separation dedicated to the cyclostationary signals, and, on the other hand, on the methods exploiting the techniques resulting from machine learning, these can be applied to different types of time signals (audios, ...). It will then be necessary to take control of existing methods to test them on already acquired cyclostationary signals. The core of the work will consist in proposing, testing and comparing one or more methods of denoising and / or separation of cyclostationary signal sources exploiting the recent methods of machine learning (deep learning, …). As part of the IMOTEP project, the tribological signal processing for the characterization of friction parts in an engine (implementation on test benches), will provide an original application domain for the implementation of these treatments, in connection with the expected progress for the industrial partners of the project (HEF-IREIS). Other applications may also be considered. Because machine learning methods may require a large number of training samples, it will be necessary to consider using and / or developing methods for the synthesis of cyclostationary signals, particularly in relation to the signals that may appear in the used test benches.

******** Candidate

We are looking for a motivated student holding an engineer diploma or a Master degree (on the 1st of September 2018) in the field of "computer science" with strong skills in applied mathematics. A good background in software development (algorithmic, Matlab/Octave/Scilab or Python, ...) is expected. Knowledges in machine learning would also be appreciated.


******** Salary

Net salary: around 1400 euros. Teaching activities are eventually possible (64 hours per year).


******** Application process

Your application should include the following documents:

- Letter of intent

- Grades and ranking during Master 1 and Master 2 (or two last years of engineer school)

- Scientific CV

- List of publications (if it exists of course)

- Names of Referees (at least 2)


******** Contacts:

- Fabien Momey <fabien.momey@univ-st-etienne.fr>

- Olivier Alata <olivier.alata@univ-st-etienne.fr> (http://perso.univ-st-etienne.fr/ao29170h/)

- Christophe Ducottet <ducottet@univ-st-etienne.fr> (https://perso.univ-st-etienne.fr/ducottet/)

******** Bibliographie


[AMX01]K. Abed-Meraim, Y. Xiang, J.H. Manton, Y. Hua, "Blind source-separation using second-order cyclostationary statistics" IEEE Transactions on Signal Processing, 49(4), 2001.

[BA97] A. Belachrouni, K. Abed-Meraim, J.-F. Cardoso, E. Moulines, "A Blind Source Separation Technique Using Second-Order Statistics" IEEE Transactions on Signal Processing, 45(2), 1997.

[GN06] W. A. Gardner, Antonio Napolitano, Luigi Paura, "Cyclostationarity: Half a century of research", Volume 86, Issue 4, April 2006, Pages 639-697.

[GP17] E. M. Grais, M. D. Plumbley, "Single Channel Audio Source Separation using Convolutional Denoising Autoencoders" In: 5th IEEE Global Conference on Signal and Information Processing (GlobalSIP2017), 14 - 16 November 2017, Montreal, Canada.

[KK18] M. Kulin, T. Kazaz, I. Moerman, E. De Poorter, "End-to-end learning from spectrum data: A deep learning approach for wireless signal identification in spectrum monitoring applications", IEEE Access, 2018. https://arxiv.org/abs/1712.03987

[KY17] M. Kolbæk, D. Yu, Z.-H. Tan, J. Jensen, "Joint Separation and Denoising of Noisy Multi-talker Speech using Recurrent Neural Networks and Permutation Invariant Training" In: IEEE Int. Workshop on Machine Learning for Signal Processing, 2017, Tokyo, Japan. https://arxiv.org/abs/1708.09588


Dans cette rubrique

(c) GdR 720 ISIS - CNRS - 2011-2018.