Research Interests
My research is in the field of audio signal processing, convex optimization and machine learning. More precisely, I developed efficient methods for restoring missing data, with a particular focus on phase reconstruction in the time-frequency domain, which is an important issue in many applications to digital audio signals. The amplitudes (spectrograms) are relatively simple to manipulate, but the phases pose specific and difficult problems. A prototypical example is the separation of audio sources. Most time-frequency approaches rely on a segmentation of the time-frequency plane, each region being associated with a "source", followed by partial syntheses of these sources from the time-frequency transform restricted to these regions. I developed a phase reconstruction procedure. The fields of applications include the restoration of old recordings or many aspects involving sound analysis and synthesis.