Master 2017 2018
Stages de la spécialité SAR
Machine Learning Algorithm for Classification of Synthesizer Sounds


Site : Arturia
Lieu : Arturia - Meylan, France (proche de Grenoble, 38)
Encadrant : Fanny Roche - fanny.roche@arturia.com
Dates :02/2018 à 08/2018
Rémunération :Indemnité de stage
Mots-clés : Parcours ATIAM : Informatique musicale, Parcours ATIAM : Traitement du signal

Cliquer ici pour vous authentifier


Description

About our Company :

Arturia is a leading developer of music software and hardware for both pro and amateur musicians around the world. Our strength and success come from our dedication to music and inspiring the people who make it. It’s our passion because many of us are musicians, as well. Founded in Grenoble, France in 1999, the company’s dedicated drive, business acumen, and operational R&D culture have earned us a leadership position among the best musical instrument manufacturers in the world. In addition to developing our own proprietary technology, we also develop strategic alliances with top innovators and research centers. In this way, we empower and expand the creative experiences of our customers in over 70 countries. In building a world-class brand, our mantras continue to be excellence in quality, and efficiency in processes. We love to design and build intuitive, elegant products that inspire, empower and streamline the artistic process for musicians.

Main Activities :

The main part of Arturia software products are virtual emulation of famous analog synthesizers and physical models of electro-acoustic instruments, in particular the Minimoog or the Jupiter8 synthesizer and the latest released, the Synclavier and the Piano modeling. Musicians can use a MIDI keyboard to control the software sound synthesis done by a software, using the computer’s own processor. Arturia also conceived a wide range of MIDI controllers, has become famous for the unique sound of the MiniBrute and MicroBrute anolog synthesizers that you can find in almost any professional or home studio, and has released hybrid products, such as AnalogLab (Synthesizers) and Spark (Drum machines), that combine MIDI Controllers with their dedicated software.

The Internship :

Arturia’s software products come with a huge database of presets (particularly AnalogLab) making it necessary for such systems to have a clear and ergonomic classification of sounds for the users to have easily access to the whole range of sounds. Until now, a sound categorization is done using manually implemented tags which are hence subjective and can make it even harder for the users to find the sound they really seek. Music Information Retrieval is a very active research field whose purpose is to extend the understanding and usefulness of music data. One of its applications is to classify musical sounds by extracting some useful descriptors from the audio signal. The aim of this internship is to work on a musical sound classification algorithm avoiding human-bias by using only objective audio characteristics. A classification algorithm using neural networks has already been implemented but need to be pushed forward and improved to be embedded in software products.   The goals for Arturia are :

  • Getting familiar with the classification method implemented
  • Studying higher-level timbre descriptors
  • Evaluating audio descriptors redundancy
  • Improving the algorithm (accuracy, efficiency)

Profile : It is necessary to have a good knowledge in Digital Signal Processing (time/frequency analysis, feature extraction) and in Machine Learning (unsupervised learning). It is also important to have some experience in C++ programming. Some skills in audio signal processing and neural networks would be a plus. Keywords : Music Information Retrieval, machine learning, audio descriptors