Job offers / Offres d'emploi

Internship / Stages



[Master]  –  Advanced DOA/DOD parametric estimation in non-Gaussian scenario: A stepwise numerical concentration approach


Recent high resolution radars, as the multiple-input multiple-output (MIMO) radar, where the scatterer number is random, exhibit non-stationarity and non-Gaussianity observations. Consequently, the common inadequate assumption of Gaussian modeling of the clutter, white or colored, deviates heavily from reality and leads to a dramatic performances loss.
Specifically, on the topic of direction-of-arrivals and direction-of-departures (DOD/DOA) estimation in a MIMO radar context, either parametric or spectral-based algorithms are usually proposed in the literature. However, these algorithms exclusively model the radar clutter as Gaussian-distributed.  On the other hand, as for estimation problems associated with non-Gaussian clutter, most of the related works deal solely with the estimation of clutter parameters, in which the texture parameter(s) and/or the speckle covariance matrix are estimated by assuming the presence of secondary data (known clutter-only realizations), instead of considering unknown clutter realizations embedded in and contaminating the transmitted/received signal. On the contrary, in order to estimate radar signal parameters under non-Gaussian clutter, a recent work done by Wang & Nehorai tackle this problem by devising parameter-expanded expectation-maximization algorithms for phased-array radar. Nevertheless, their proposed algorithms are restricted to a special, linear signal model (called the generalized multivariate analysis of variance -GMANOVA) model) thus cannot be directly applied to general MIMO radar models, nor to the DOD/DOA estimation problems, which are highly nonlinear and of interest to the practitioner.
To sum up, and to the best of our knowledge, there exists no available algorithm in the current literature that addresses the DOD/DOA estimation problem in the presence of a realistic non-Gaussian clutter. 
The aim of this internship is to fill this gap by developing new accurate and computationally efficient method for target localization in a non-Gaussian scenario in the context of co-located MIMO radar and assessing and quantifying their performances. This topic is very timely due to the increase usage of high resolution radars for which the practical interest is more than justified. The so-called complex elliptically symmetric (CES) distribution, thanks to its ability to describe different scales of the clutter roughness and to incorporate various non-Gaussian distributions, becomes recently a commonly used modelling especially in the MIMO radar context. Consequently, we will adopt this modelling. Specifically, we will tackle the design of such algorithms based on the so-called stepwise numerical concentration approach of the maximum likelihood estimator which is recently shown to be an efficient tool in order to design fast and nearly optimal schemes. A performance analysis of the proposed estimator will be investigated through existence and uniqueness of solutions, invariance to different speckle covariance matrix assumptions, convergence and computational cost analysis.


References:
[1] X. Zhang, M. N. El Korso and M. Pesavento, ''MIMO radar target localization and performance evaluation under SIRP clutter'', Signal Processing Journal, Elsevier, Volume 130, January 2017, Pages 217–232
[2] E. Ollila, D. E. Tyler, V. Koivunen, and H. V. Poor, “Complex elliptically symmetric distributions: Survey, new results and applications,” IEEE Trans. Signal Processing, vol. 60, pp. 5597–5625, Nov. 2012
[3] C. Ren, M. N. El Korso, J. Galy, E. Chaumette, P. Larzabal and A. Renaux, "On the Accuracy and Resolvability of Vector Parameter Estimates", IEEE Transactions on Signal Processing, Volume: 62, Issue: 14, Jul. 2014, pp. 3682-3694
[4] M. Greco, Y. Abramovich, J.P. Ovarlez, H. Li and X. Yang, ”Introduction to the Issue on Advanced Signal Processing Techniques for Radar Applications”, Selected Topics in Signal Processing, IEEE Journal of, 9(8), pp.1363-1365, 2015, DOI: 10.1109/JSTSP.2015.2497458
[5] M. Mahot, P. Forster, F. Pascal and J.P. Ovarlez, ”Asymptotic Properties of Robust Covariance Matrix Estimates”, Signal Processing, IEEE Transactions on, 61(13), pp.3348-3356, July 2013, DOI: 10.1109/TSP.2013.2259823
[6] C. Ren, M. N. El Korso, J. Galy, E. Chaumette, P. Larzabal and A. Renaux, "Performance bounds under misspecification model for MIMO Radar application'', in Proc. EUSIPCO 2015, Nice, France, August 2015.
[7] M. Haardt, M. Pesavento, F. Röemer and M. N. El Korso, "Subspace Methods and Exploitation of Special Array Structures", Electronic Reference in Signal Processing: Array and Statistical Signal Processing, vol. 3, pp. 651-717, Academic Press Library in Signal Processing, Elsevier Ltd., 2014, Chapter 2.15, ISBN 978-0-12-411597-2.
[8] X. Zhang, M. N. El Korso and M. Pesavento, ''Maximum likelihood and maximum a posteriori direction-of-arrival estimation in the presence of SIRP noise'', in Proc. IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP 2016), Shaingai, China, March 2016.
[9] M. Brossard, M. N. El Korso, M. Pesavento, R. Boyer, P. Larzabal and S. J. Wijnholds, "Parallel Calibration for Sensor Array Radio Interferometers", https://arxiv.org/abs/1609.02448.

Conditions:
- An outstanding and highly motivated candidate is solicited in the SONDRA laboratory of CentraleSupélec (Campus Gif-Sur-Yvette) for a 6 months internship.
- Students at the M.Sc., M. Eng., Master (or equivalent) level with background in signal processing, probability, statistics, or optimization theory are encouraged to apply.
- Good mathematical background is required. Above all, the applicants must be motivated to learn quickly and work effectively on challenging research problems.


Application process:
Please send your CV, motivation letter, transcripts of grades with qualifications and pertinent information, as soon as possible, to Dr. J.-P. Ovarlez (jeanphilippe.ovarlez@supelec.fr), Dr. C. Ren (chengfang.ren@centralesupelec.fr) and Dr. M. N. El Korso (m.elkorso@u-paris10.fr)


[Master]  – Effects of resolution degradation on the low frequency signature of an urban canyon

Previous studies have been conducted to investigate the use of multipath propagation between neighboring buildings in order to reach and detect a target hidden in the urban canyon. It was shown that at lower frequencies, the scattering mechanisms change due to the wider aperture of the incident and reflected beam. As a result, the ground of the canyon is entirely illuminated and the intensity of the surface currents remains sufficiently high for detection. As the previous study was done with very high resolution, the aim of this Master is to study the low frequency signature for lower and more realistic resolutions.


=> Contact


[Master+PhD] NRC – Distributed SAR



=> Contact + Subject




PhD position / Thèses


[PhD] PSCM – Modeling and characterization of human signature under foliage

The goal of this PhD is to investigate the feasibility of detection of human motion in a FoPen (Foliage Penetration) configuration. The novelty of the proposed research work is on the analysis of micro-Doppler signature at low frequency (UHF) in a context of FoPen detection. Due to the complexity of the modeling in this frequency band, an experimental approach would be proposed, in conjunction with simulations fed by the signatures measured.


=> Contact


[Master + PhD] PSCM – 3D modeling of EM propagation through and over urban areas at low frequencies

Passive radars exploit illuminators of opportunity to detect targets. These illuminators, e.g. GSM, FM and TV broadcast signals, are typically in the lower frequencies and are pervasive in urban areas. Thus, there are possibilities of using passive radars to detect low-flying, low-speed targets within and over urban areas. In this PhD, the student will study the effects of EM propagation through and over urban areas for the purpose of radar system budget analysis and signal processing analysis. As this is a complex topic, Master project is also planned to clearly define areas to focus on for the PhD.


=> Contact



[PhD] NRC – Model-based classification for urban targets detection

Due to their small size and limited cruising speed, detecting and identifying multi-rotors is quite challenging. Their small size often requires changing the radar thresholds and sensitivity to be able to gather enough energy back for the processing. By doing so, many other small objects (birds for example) can trigger the detection process of the radar. To estimate properly the nature of the detected object, more knowledge of the dynamic response of the UAV to the radar signals is required. This PhD aims at studying the physical interactions between electromagnetic waves coming from the radar and the different mechanical parts of the UAV during its flight. This study should be driven to build a realistic model of the fluctuations of radar echoes, so that it could be used for UAV classification or identification.


=> Contact



[PhD] – Activity monitoring using time series of SAR or EO or both types of images

In the last decade, many new satellite remote sensing missions (for example Sentinel-1) have been launched, resulting in dramatic improvement in the image acquisition capabilities. This increases the availability of multi-temporal and muti-dimensional images of the Earth surface, with improved temporal and spatial resolution, hence the interest of time series processing. The aim of this PhD is to work on new techniques for detecting and monitoring changes in an area under surveillance using a time series of SAR or EO images or a combination of both. The key requirement is high probability of detection and low probability of false alarms. The time scale of changes could range from hours to days to months. One important application is key installation monitoring.

=> Contact

[PhD]  – Active meta-surface for scanning antennas

In this PhD, we consider an array of resonators (Electric-field-coupled resonator for example) whose resonant frequency may be tuned through the use of biased elements such as varactor diodes or liquid crystals. Tunable elements are arranged in a precisely calculated pattern. Radiofrequency (RF) energy is scattered when the elements are activated, holographically generating a beam. The direction of the beam is defined by the specific elements that are electronically activated. We can therefore expect to have three different modes with the same antenna: continuous tracking mode, instantaneous switching mode, and beam lock mode. This approach could lead to light and electronically simple scanning antenna (avoiding complex phase shifting modules of large antenna array) that can be used for communications for radars for small UAV, etc.

=> Contact








Post-Doc










 
Dernière modification : 21/11/2016