Description
Date depot: 1 janvier 1900
Titre: Probabilistic latent variable analysis for automatic music transcription
Directeur de thèse:
Axel ROEBEL (STMS)
Domaine scientifique: Sciences et technologies de l'information et de la communication
Thématique CNRS : Non defini
Resumé:
{{Introduction}}
The automatic transcription of a piece of music consists of deriving a midi score that contains sufficient information to replay the original piece of music with a midi synthesiser. Recent research results have allowed making considerable progress towards this goal without however coming close to a complete solution.
A multitude of algorithms are available that solve the important underlying sub-problems: polyphonic pitch estimation (Yeh, et al., 2010; Dressler, 2014), and note onset detection (Roebel, 2003; Boeck, et al. 2012). And recently a new probabilistic interpretation for nonnegative decomposition emerged und the term termed probabilistic latent component analysis (PLCA) (Shashanka et al. 2008). The probabilistic interpretation allows for straightforward integration of prior information and already has been used for automatic transcription with very promising results in (Benetos, 2012; Grindlay et al; 2015).
Doctorant.e: Jacques Celine