Olivier Besson - Research

Sommaire

Research area

Olivier Besson research activities are in the general area of statistical signal and array processing with particular interest to robustness issues in detection/estimation problems for radar and communications. Lines of research include :

robust adaptive beamforming
array calibration techniques
adaptive radar detection with spatial or space-time signature uncertainties
space-time adaptive processing in heterogeneous clutter
adaptive detection in low-sample support
wideband radar processing
space division multiple access in satellite communication systems.

see below for a technical description and here for a list of publications.

Cooperations

Industry and governmental agencies : Thales Alenia Space, Thales Airborne Systems, ONERA, DGA.
Academia : Università del Salento (F. Bandiera, G. Ricci), ENSEEIHT (J.-Y. Tourneret), Colorado State University (L.L. Scharf), Uppsala University (P. Stoica), Darmstadt University (A. Gershman), MIT Lincoln Laboratory (S. Kraut).

Professional activities

Associate Editor for the IEEE Transactions Signal Processing
Member of the Sensor Array and Multichannel technical committee (SAM TC) of the IEEE Signal Processing Society
Regular reviewer for IEEE Transactions (SP, AES, Comm) and conferences (ICASSP, SAM, SSP)

Research overview : A generic detection problem we are interested in can be described by the following binary composite hypothesis testing problem

$\begin{align*} H_{0}: & \begin{cases} \boldsymbol{z} = \boldsymbol{n}; \\ \boldsymbol{z}_l = \boldsymbol{n}_k; \; k=1,\cdots,K \end{cases} \nonumber \\ H_{1}: & \begin{cases} \boldsymbol{z} = \alpha \boldsymbol{v} + \boldsymbol{n}; \\ \boldsymbol{z}_k = \boldsymbol{n}_k; \; k=1,\cdots,K \end{cases} \end{align*}$

where $\boldsymbol{z}$ is the received signal and $\boldsymbol{v}$ is the assumed space and/or time signature of the signal of interest (SOI) whose presence we wish to detect ( e.g. a target). $\boldsymbol{n}$ stands for the noise vector whose covariance matrix is $\boldsymbol{R}=\mathcal{E} \left\{ \boldsymbol{n} \boldsymbol{n}^{H} \right\}$ and $\boldsymbol{n}_{k}$ are training samples with covariance matrices $\boldsymbol{R}_{k}=\mathcal{E} \left\{ \boldsymbol{n}_{k} \boldsymbol{n}_{k}^{H} \right\}$ . Ideally, $\boldsymbol{R}_{k}=\boldsymbol{R}$ and $\boldsymbol{v}=\boldsymbol{s}$ where $\boldsymbol{s}$ is the actual signature of the SOI. We consider situations where either one or both hypotheses are no longer fulfilled, and study detectors that can handle such mismatches.

A closely related problem is beamforming where, from a set of snapshots $\boldsymbol{X}=\begin{bmatrix} \boldsymbol{x}_1 & \boldsymbol{x}_2 & \cdots & \boldsymbol{x}_K \end{bmatrix}$ which consist of interference, noise (and possibly the SOI), one wishes to design a filter $\boldsymbol{w}$ that eliminates noise while letting the SOI pass. This typically amounts to solving

$\min_{\boldsymbol{w}} \boldsymbol{w}^H \boldsymbol{R} \boldsymbol{w} \: s.t. \boldsymbol{w}^H \boldsymbol{v}=1$

where $\boldsymbol{R}=\mathcal{E} \left\{ \boldsymbol{X} \boldsymbol{X}^{H} \right\}$ and $\boldsymbol{v}$ is the SOI steering vector. Again, we address cases where there exist uncertainties about $\boldsymbol{v}$ , where $\boldsymbol{X}$ could include the SOI and few snapshots are available.

Back to homepage

Research

Research area

Cooperations

Professional activities

Navigation

Département

Groupe de recherche

Axe(s) de recherche

Unité(s) de formation