Blind Source Separation: Sparse Component Analysis for Convolutive Mixtures and Nonlinear Mixtures

• Grant number: 15/07048-4
• Support Opportunities: Scholarships in Brazil - Post-Doctoral
• Effective date (Start): September 01, 2015
• Effective date (End): January 24, 2017
• Field of knowledge: Engineering - Electrical Engineering - Telecommunications
• Principal Investigator: João Marcos Travassos Romano
• Grantee: Kenji Nose Filho
• Host Institution: Faculdade de Engenharia Elétrica e de Computação (FEEC). Universidade Estadual de Campinas (UNICAMP). Campinas , SP, Brazil

Abstract

Recently, sparse component analysis has become one of the most powerful tools for the blind source separation problem, under the hypothesis that the sources are sparse in a given domain. The multidisciplinary characteristic of this problem and its comprehensive formulation allows its applicability in several areas of interest such as hyperspectral images, speech, audio, seismic reflection, chemical sensors, biomedical signals and communications.In his PhD thesis, the candidate has been working on relevant contributions in the processing of signals with a certain degree of sparsity, with emphasis on reflection seismic applications. For that, it has been proposed new methods for the deconvolution and blind source separation problems. However, the proposed methods are limited to structurally simpler models such as the mono channel convolution case and the linear instantaneous mixture model.As a natural evolution of his work, this postdoctoral plan of work deals with the blind source separation of more structurally complex models in the context of sparse component analysis. More specifically, this plan of work deals with a particular case of convolutive mixture, the multichannel convolution, and two different nonlinear mixture models: the linear-quadratic and the post-nonlinear mixture models.

Scientific publications

(References retrieved automatically from Web of Science and SciELO through information on FAPESP grants and their corresponding numbers as mentioned in the publications by the authors)

NOSE-FILHO, KENJI; TRAVASSOS ROMANO, JOAO MARCOS. Low-Rank Decomposition Based on Disjoint Component Analysis With Applications in Seismic Imaging. IEEE TRANSACTIONS ON COMPUTATIONAL IMAGING, v. 3, n. 2, p. 275-281, JUN 2017. (15/07048-4)