Directeur adjoint
Professeur des universités
Publié le – Mis à jour le
Thématiques de recherche : My research interests are divided in three categories.
Originally, I am a specialist of musical sound modeling, with applications to music industry and creation. More precisely, I am an expert of pseudo-periodic signals and time-frequency representations, applied to the analysis, retrieval, or separation of sound signals.
I also work on image analysis and computer vision, with applications to cultural heritage (ancient coins). I am particularly interested in the human perception of sounds and images.
More recently, I consider other types of sensors (not only ears or eyes) for computers to take decisions in the context of autonomous vehicles (intelligent trains, semi-autonomous sailing boats).
Points forts des activités de recherche : I am a pionneer of sinusoidal modeling for computer music, or spectral modeling,and I proposed a method for the precise estimation of sinusoids (spectral analysis) based on the Fourier transform,using derivatives of the signal, as well as a method for the tracking of these sinusoids in time, based on linear prediction.
Sinusoids indeed correspond to spectral peaks that turn out to be of great interest for e.g. source separation or music retrieval. Moreoever, they show great advantages for physics (energy), mathematics and computer science (sparsity), and psychoacoustics (perception).
Regarding source separation, I am at the origin of the informed source separation (ISS) approach, which consists in using extra information to enhance the quality of the solutions proposed for inverse problems.
Regarding music retrieval, I have contributed to the perceptual hashing of music pieces (see the Shazam system) but also of images.
Currently, I am also starting an activy of analysis of natural scenes (sea), in the context of autonomous vehicules (sailing boats), with embeded sensors and computing. The common link between these contributions is to build a chain from sensors to knowledge then decision.
Page perso : https://www.sylvain-marchand.info/
