In the Music Technology Group I've done research on Music Information Retrieval on topics like;
music similarity, music performance analysis, and sound quality analysis. This research is been carried out
from a data-driven perspective.
Good-sounds
Good-sounds.org has been the result of a research collaboration between KORG Inc. and the Music Technology Group.
The project was meant to develop technologies for automatic sound quality assessment.
I created the data collections that were openly published as the
good-sounds dataset. With this data
I created machine learning models to automatically evaluate instrumental sounds in several dimensions.
We also built a website (good-sounds.org, no longer accessible) in which
we created a community of users that upload and annotate sounds to help us improve the models we
previously built with our own data.
Out of this collaboration, Korg released the mobile app
Cortosia and the
TM-50TR tuner.
We also published two research articles in the main conferences in the field:
Bandiera et al., 2016 and
Romaní Picas et al., 2015.
Music Critic
Music critic is the product of the
TECSOME
project, which is funded by a Proof of Concept European Research Council grant.
The project came up from an idea of my own, which was to build a web API to evaluate music exercises that would be easily integrated in online education platforms.
I implemented the application and we researched on music similarity using machine learning techniques to build the models for the automatic evaluation of the
music performances.
The first online music course to use the technology was developed in the online education platform
Kadenze .
We also published some papers, like
Bozkurt et al., 2018