My thesis is at the crossroads of statistical signal processing, tensor methods, and aeroacoustics , with a focus on aircraft engine noise (funded by Safran Tech). I develop methods to separate broadband acoustic sources in turbofan noise (fan, jet, core, combustion), modeled as centered, independent Gaussian processes, from microphone array measurements on engine test benches. The first line of work introduces informed nonnegative matrix factorization for realistic settings where only auto-spectra from previous test campaigns are available, encoding prior knowledge on turbomachine noise through spectral templates and parametric constraints. The second line of work proposes a positive semi definite tensor factorization of spatial covariance matrices as a multichannel extension of NMF, exploiting the low rank structure of latent sources and all spatial covariance information under high background noise and unknown propagation conditions.
Other than that, I enjoy biking to : work, vallée de chevreuse, and basically anywhere. I am always glad to lend a hand if your bike needs to be fixed.
My supervisors are Gilles Chardon and Charles Soussen. They are great.
