Banca de DEFESA: CAROLINE PIRES ALAVEZ MORAES

Uma banca de DEFESA de DOUTORADO foi cadastrada pelo programa.
STUDENT : CAROLINE PIRES ALAVEZ MORAES
DATE: 27/05/2024
TIME: 14:00
LOCAL: https://meet.google.com/iuz-ictv-ouj
TITLE:

Multifaceted IVA: New Approaches for BCI Applications

PAGES: 102
BIG AREA: Engenharias
AREA: Engenharia Elétrica
SUMMARY:

Joint Blind Source Separation (JBSS) is an extension of the Blind Source Separation (BSS) problem to multiple datasets, being an important research topic applied in many areas due to its wide versatility. Dealing with several datasets and exploring their information simultaneously has called the attention of researchers in the last decade and is still a challenge. Among the several possible methods in the context of JBSS, the Independent Vector Analysis (IVA) is an interesting approach, since it is based on the Independent Component Analysis (ICA) and it explores the statistical dependency between different datasets through the use of Mutual Information (Mutinf). This work proposes three different approaches of IVA to deal with biomedical signals, focused on the Motor Imagery (MI) paradigm and a possible extension to Epilepsy Seizures disorder. Firstly, in the context of MI classification based on Electroencephalogram (EEG) signals for Brain-Computer Interface (BCI), several methods have been proposed to extract features efficiently, mainly based on common spatial patterns and filter banks. In this scenario, we propose an original approach for feature extraction in motor imagery, based on exploring the minimization of mutual information through IVA followed by the use of autoregressive filters. For the classification of imaginated movements, we used consolidated classifiers: Linear Discriminant Analysis, Support Vector Machines, Knearest neighbors, and deep classifiers: EEGNet and EEG-Inception. This approach was evaluated in two different MI datasets: BCI Competition IV - Dataset 1 (DS1) and BCI Competition III - Dataset 4a (DS4a). In the second approach, we propose an innovative application of IVA as a Transfer Learning technique for MI classification. The statistical dependency between datasets exploited through Mutinf could aid in MI classification since it allows a generic and homogeneous treatment of the whole data and a possible knowledge transfer between subjects/patients. The results were evaluated in the DS1 dataset, showing a high correlation and small standard deviation between cross-subjects. Inspired by the previous results, in the third approach we present the results for the clustering technique with an epileptic dataset that was developed in collaboration with the Signal Processing Laboratory 4 (LTS4) in École Polytechnique Federal de Lausanne (EPFL), in Switzerland.

COMMITTEE MEMBERS:
Presidente - Interno ao Programa - 1544392 - ALINE DE OLIVEIRA NEVES PANAZIO
Membro Titular - Examinador(a) Interno ao Programa - 1761107 - RICARDO SUYAMA
Membro Titular - Examinador(a) Interno ao Programa - 1946319 - DIOGO COUTINHO SORIANO
Membro Titular - Examinador(a) Externo à Instituição - SARAH NEGREIROS DE CARVALHO LEITE - ITA
Membro Titular - Examinador(a) Externo à Instituição - RAFAEL FERRARI - UNICAMP
Membro Suplente - Examinador(a) Externo ao Programa - 1676329 - RAPHAEL YOKOINGAWA DE CAMARGO
Membro Suplente - Examinador(a) Externo à Instituição - LEONARDO TOMAZELI DUARTE - UNICAMP