Postdoctoral researcher M/F
Duration: 18 months
Location: to be discussed with the successful candidate, either IREENA (Saint-Nazaire, Univ. Nantes) or CRIStAL (Univ. Lille)
The ANR project ASCETE had its kickoff in early 2020. It is devoted to new approaches in time-frequency analysis, motivated by the analysis and separation of complex signals. In this postdoc, we will focus on a recent approach to time-frequency filtering using the zeros of spectrograms, see [Fla15]. While traditional approaches to filtering focus on the local maxima of a spectrogram, the zeros seem to have more mathematical structure, and there are hints that they could be more robust to lower signal-to-noise ratios than maxima.
The first task of the postdoc will be to set up a reproducible experimental benchmark to test such claims. This will serve as an introduction to some of the tools and techniques related to zeros. The latter cover a wide range of domains in a blend characteristic to signal processing, including computational harmonic analysis, spatial statistics, and stochastic processes, see e.g. [Bar19]. At this stage, the second task will be determined by the whole team, depending on the tastes, skills, and objectives of the postdoc. There is a rich set of questions to be addressed on zeros of spectrograms, and new research questions are welcome.
The location of the position is to be chosen among
1)IREENA, where François Auger is based.
2)CRIStAL, the research centre in computer science, signal processing, and automatic control of Lille, where Rémi Bardenet and Pierre Chainais are based.
In any case, frequent visits between Nantes-Saint-Nazaire, Lille, Lyon, and other sites who participate in the Ascete project (Grenoble, Paris) will be possible.
[Fla15]Flandrin, P. Time–frequency filtering based on spectrogram zeros. IEEE Signal Processing Letters, 22(11):2137–2141, 2015.
[Bar19] R. Bardenet, J. Flamant, P. Chainais, “On the zeros of the spectrogram of white noise”, Applied and Computational Harmonic Analysis (Elsevier), Vol 48, No 2, 2020.
(c) GdR 720 ISIS - CNRS - 2011-2020.