Design, implementation and evaluation of deconvolution methods in periodic biomedical signals
Design, implementation and evaluation of deconvolution methods in periodic biomedical signals
Date
Authors
Μακρυγιώργου, Δήμητρα
Journal Title
Journal ISSN
Volume Title
Publisher
Abstract
In the present thesis, an extensive study and simulation of source separation methods for periodic biosignals are conducted. Biosignals nature and the emerged from their origin and acquisition characteristics make the separation a cumbersome task. A wide variety of source separation techniques have been implemented, each one based on a different assumption, relative to signals characteristics. Methods based on signals correlation, independency, and periodicity are studied. Specifically, fastICA, Infomax, JADE, AMUSE, SOBI, πCA, natural gradient methods and contrast function approaches for convolutive mixture and algorithm based on signal cyclostationarity are simulated and examined on both synthetic data and real biosignals. Finally, an approach, that takes into account the frequency components of the under investigation signals and its inputs are one channel data, is implemented and examined on real biosignals.
Description
Keywords
Source separation, Biosignals, Instantaneous and convolutive mixtures, Independent component analysis, Periodic component analysis, Cyclostationary methods