I’m a PHD student in SCREAM lab, NCKU, Taiwan. Before I came to IRCAM, my research was note separation based on recursive-regression non-negative factorization(RR-NMF). Due to have a better analysis directly from the real performance recording, we want to develop a better note separation not only for the sources, but also for the notes in a source. The result of note separation can be applied on the audio analysis, resynthesize and audio information retrieval.
Last July (2017), I got a scholarship for one year and came to IRCAM. The research topic is music transcription based on convolutional neural network(CNN). Although, recurrent neural network(RNN) is more usually used for time continuity signal, for music transcription also can be processed with short term signal analysis. Therefore, recently, there are more and more researchers also use CNN for audio signal analysis
Courriel : yi.ju.lin (at) ircam.fr
- Équipe : Analyse et synthèse des sons