Study and application of blind source separation algorithms

• Grant number: 17/06974-8
• Support Opportunities: Scholarships in Brazil - Scientific Initiation
• Start date: July 01, 2017
• End date: June 30, 2018
• Field of knowledge: Engineering - Electrical Engineering - Telecommunications
• Principal Investigator: Levy Boccato
• Grantee: Leonardo Fill Cardoso
• Host Institution: Faculdade de Engenharia Elétrica e de Computação (FEEC). Universidade Estadual de Campinas (UNICAMP). Campinas , SP, Brazil

Abstract

The problem of blind source separation consists in extracting several signals (sources) from the set of collected observations, which are composed of mixtures of the signals thereof. The term blind indicates that there is a minimum amount of available information about the mixing system and/or about the sources, which makes the problem quite challenging. A well-established approach for source separation is based on the hypothesis that the original signals are statistically independent. Such perspective, known as Independent Component Analysis (ICA), aims at creating a separation system capable of generating estimates of the sources that are as independent as possible. In order to implement this strategy, it is necessary to define a mathematical criterion that expresses the notion of independency, as well as to design an efficient algorithm to adjust the parameters of the separation system in order to optimize the cost function associated with the adopted criterion. The objective of this research project is to study the problem of blind source separation along with the methodologies that can be used to solve it, especially those related to the ICA approach. We shall investigate the different existing criteria for ICA, such as those based on mutual information and non-gaussianity, and, then, we shall implement the main algorithms for source separation, which shall be applied in different scenarios, considering both audio and image sources, mixing systems containing noise and/or of difficult inversion, in order to assess the performance of each technique and to establish a comparative analysis between the studied algorithms. (AU)