Separation of sounds of musical instruments

• Grant number: 15/25992-1
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: April 01, 2016
• End date: March 31, 2017
• Field of knowledge: Engineering - Electrical Engineering - Telecommunications
• Principal Investigator: Magno Teófilo Madeira da Silva
• Grantee: André Oliveira Françani
• Host Institution: Escola Politécnica (EP). Universidade de São Paulo (USP). São Paulo , SP, Brazil

Abstract

Techniques of blind source separation (BSS) have attracted attention of the scientific community in the recent decades. They can be applied in the processing of different kind of signals, as signals of a multiuser communication system, biomedical signals, astronomy/cosmology, geophysical, speech, music, among others. Recently, these techniques have been applied to the separation of sounds of musical instruments and also to the separation of the singer's voice and the sounds of the instruments present in a recording. This application represents a challenge in the context of source separation, since musical sources are highly correlated in time and frequency. In this project, we intend to study BSS techniques to separate sounds of different musical instruments. We will address the classical source separation methods based on the independent component analysis and also those based on nonnegative matrix factorization taking into account the score. These methods will initially be applied to mixtures of synthesized sounds of musical instruments. Therefore, we will also revisit techniques of musical synthesis in the project. Finally, we intend to apply the studied techniques to not synthesized mixtures.