Advanced search
Start date

Study of sparsity-promoting regularization for superdirectional microphone with spherical geometry

Grant number: 18/26531-6
Support type:Scholarships in Brazil - Master
Effective date (Start): March 01, 2019
Effective date (End): February 28, 2021
Field of knowledge:Engineering - Electrical Engineering - Telecommunications
Principal Investigator:Bruno Sanches Masiero
Grantee:Fernanda Caldas Correia
Home Institution: Faculdade de Engenharia Elétrica e de Computação (FEEC). Universidade Estadual de Campinas (UNICAMP). Campinas , SP, Brazil
Associated research grant:17/08120-6 - 3D-Audio. Sampling, encoding and reproduction of spatial audio, AP.JP


The determination of the quantity and direction of the sources present in a sound scene can be made, in principle, using spatial filters to estimate the sound level arriving at a microphone arrangement from different directions, similar to the estimation of acoustic images . This technique, however, usually presents low resolution due to the relatively small number of microphones constituting the arrays. Several methods to increase the quality of the acoustic image estimation without increasing the number of sensors have been proposed. A more recent alternative is the use of sparse-promoting regularization, which uses the 1 norm in solving the problem of determining sound objects. These algorithms work giving greater weight to sparse scenes (which is usually a valid assumption for scenes made up of a small number of sound sources), promising a better separation of the present objects. The techniques mentioned above are usually applied for planar arrays. However, planar arrays allows the detection of sources only in one hemisphere. To allow the search for sources in all directions, it is necessary to use arrangements with a three-dimensional geometry, the spherical arrangement being the most commonly used geometry. This project aims to study sparsity promoting methods for detection and separation of sound objects using spherical microphone arrays.