Passer au contenu

Armando Muscariello

Document Actions
Armando Muscariello, METISS Project/Team at INRIA/IRISA

IRISA-INRIA Rennes Bretagne Atlantique

Campus Universitaire de Beaulieu

263 Avenue du General Leclerc

mailto:  amuscari(at)

Phone number:  +33 (0) 2 99 84 74 40

Short Bio

I was born on January 10th, 1983 in Sassari, Sardinia, Italy. I grew up over there before moving to Turin to pursue degrees at Politecnico di Torino, in Telecommunication Engineering. I completed studies on January 2007, and spent some months working as an assistant researcher  (or assegnista di ricerca, following the Italian nomenclature), together with Alessandro Nordio and Carla Fabiana Chiasserini.  The focus on the work, as well as of my Mater thesis, consisted in evaluating the quality of the reconstruction of band-limited multi-dimensioned fields from non equally spaced samples.

In November 2007 I started a doctorate in IRISA-INRIA Rennes Bretagne Atlantique under the supervision of Guillaume Gravier and Frédéric Bimbot, which I successfully defended on January 25th, 2011. The title of the dissertation is "Variability tolerant discovery of arbitrary repeating patterns in audio data with template matching" (I'll upload the manuscript as soon as the very last modifications will be done...yes, after the actual defense). During my doctoral studies, I designed and implemented a computational architecture for performing unsupervised audio motif discovery, that is the task of discovering and collecting occurrences of repeating patterns in audio streams. Example of audio motifs targeted are repeating words in speech data, or much longer (and less variable) ones like songs, jingles, advertisements in broadcast streams. Main challenge raised from the radical level of unsupervision that was assumed, Unlike more traditional train and test paradigm based on modelling, training data, and supervised machine learning techniques, we tried to perform discovery by direct pattern matching, in the absence of linguistic (in the speech case), or acoustic knowledge (in the form of training data or models).