Creating and sharing knowledge for telecommunications

Project: Dependent Component Analysis of Biological Signals

Acronym: DECA-Bio
Main Objective:
This project is meant to support a three-way collaboration between IT-Lisbon, IT-Coimbra, and Aalto University in Helsinki, Finland. The three research groups involved
have two things in common: they all work in blind source separation (BSS) in situations where the typical approach through Independent Component Analysis (ICA) is not appropriate; and they all apply these techniques to biological signals.

Our proposal is to pool together the e orts of the three research groups. They have recently worked in two main domains: separation of phase-locked subspaces; and separation of cardiac sounds. Although these two domains are very di erent, they share several aspects: 1) they are both inverse problems; 2) the observation model is linear; 3) the sources are assumed to be statistically dependent.

The team will develop novel algorihms to tackle these two separation problems, based mainly on two approaches: matrix factorization and Bayesian inference.
Reference: 0000
Funding: IT/LA
Start Date: 01-06-2011
End Date: 01-05-2013
Team: José Manuel Bioucas Dias, Miguel Tavares Coimbra, Ricardo Nuno Vigário, Miguel Sousa Borges de Almeida, Fabio de Lima Hedayioglu
Groups: Pattern and Image Analysis – Lx
Local Coordinator: José Manuel Bioucas Dias
Links: Internal Page
Associated Publications