Advanced search
Start date
Betweenand


Novas metodologias de aprendizado baseado na Teoria da Informação para equalização adaptativa

Full text
Author(s):
Denis Gustavo Fantinato
Total Authors: 1
Document type: Doctoral Thesis
Press: Campinas, SP.
Institution: Universidade Estadual de Campinas (UNICAMP). Faculdade de Engenharia Elétrica e de Computação
Defense date:
Examining board members:
Romis Ribeiro de Faissol Attux; Magno Teófilo Madeira da Silva; Charles Casimiro Cavalcante; João Marcos Travassos Romano; Rafael Ferrari
Advisor: Romis Ribeiro de Faissol Attux; Aline de Oliveira Neves Panazio
Abstract

In signal processing, statistically dependent signals carry valuable information to solve problems of various natures. Based on the classical second-order statistical framework, however, their statistical characterization is limited to a certain degree of partiality. In view of this, in this work, a more extensive extraction of the information regarding statistical dependence is proposed via the use of methods based on Information Theoretic Learning (ITL) allied to a multivariate perspective. Focusing on the statistical temporal dependence, this approach is applied to three important problems within the signal processing area: blind channel equalization with temporally-structured sources, supervised equalization using Infinite Impulse Response (IIR) filters, and nonlinear Blind Source Separation (BSS) problems. In each case, the results led to relevant contributions, including the extension of the ITL paradigm to the multivariate perspective and also to the use of metaheuristics as optimization strategies, instead of the traditional gradient-based methods. The developed study opens new possibilities for the statistical processing of videos, images and speech data in complex scenarios; in communications, it becomes possible to deal with messages subject to statistically dependent coding schemes (AU)

FAPESP's process: 13/14185-2 - New Methods for Adaptive Equalization Based on Information Theoretic Learning
Grantee:Denis Gustavo Fantinato
Support Opportunities: Scholarships in Brazil - Doctorate