Banca de DEFESA: DOUGLAS CANONE GARCIA

Uma banca de DEFESA de DOUTORADO foi cadastrada pelo programa.
STUDENT : DOUGLAS CANONE GARCIA
DATE: 01/08/2023
TIME: 14:30
LOCAL: https://conferenciaweb.rnp.br/webconf/francisco-156
TITLE:

Non-Invasive Electrophysiology Biosignal Classification (sEMG and EEG) in the Context of Hybrid Brain-Computer Interfaces (hBCI)

PAGES: 118
BIG AREA: Engenharias
AREA: Engenharia Biomédica
SUBÁREA: Bioengenharia
SPECIALTY: Processamento de Sinais Biológicos
SUMMARY:

With the aim of developing interfaces that allow greater independence for people affected by motor limitations, the concept of Brain-Machine Interface (BCI) was developed, which proposes the development of assistive technology devices using the analysis of signals produced by the brain. Motivated by the limitations of the BCI, the concept of hybrid Brain-Machine Interface (hBCI) was developed, where through the fusion between brain signals and biosignals obtained from other parts of the body, it is possible to compose a hybrid system with the potential to maximize command classification results. This thesis is based on the review of articles related to hBCI advances in the last 12 years, presenting a multifaceted research scenario where researchers develop algorithms and present solutions that are difficult to reproduce due to the low level of database sharing. With the objective of contributing to the development of a more prolific research scenario within the scope of hBCI, in this work two public databases were used: Multimodal Signal Dataset (MMSD), where EEG and sEMG biosignals are recorded in synchrony with the execution of previously established movements, and BCI Competition IV (dataset 2a), used for validation. The main scientific contribution of this thesis is the development of a same and unique biosignal processing pipeline, with the potential to classify both electroencephalographic (EEG) and electromyographic (sEMG) signals, captured (non-invasively) respectively on the surface of the scalp and the skin. The processing pipeline developed in this thesis makes use of a biosignal feature extraction system called FBRCSP (Filter Bank Regularized Common Spatial Pattern), which allows the selection of biosignal features from EEG and sEMG through the NCA (Neighborhood Component Analysis) technique, which in turn enables the classification of biosignals through the Support Vectors Machine (SVM) technique, where Accuracy and Cohen's Kappa classification metrics are used for performance evaluation. The results obtained for the EEG signals from the BCI Competition IV database (dataset 2a) surpass the classification results achieved through the use of classical machine learning approaches, published in scientific journals in the last 12 years. The same processing pipeline was applied to the sEMG signals from the MMSD database, allowing the classification of groups of movements with an average accuracy between subjects of 84.5%, and an average Kappa of 0.722.

COMMITTEE MEMBERS:
Presidente - Interno ao Programa - 1545987 - FRANCISCO JOSE FRAGA DA SILVA
Membro Titular - Examinador(a) Interno ao Programa - 1672975 - JOAO RICARDO SATO
Membro Titular - Examinador(a) Interno ao Programa - 1544392 - ALINE DE OLIVEIRA NEVES PANAZIO
Membro Titular - Examinador(a) Externo à Instituição - LEIA BERNARDI BAGESTEIRO - SFSU
Membro Titular - Examinador(a) Externo à Instituição - TIAGO OLIVEIRA WEBER - UFRGS
Membro Suplente - Examinador(a) Interno ao Programa - 1761107 - RICARDO SUYAMA
Membro Suplente - Examinador(a) Externo ao Programa - 1955999 - ANDRE MASCIOLI CRAVO
Membro Suplente - Examinador(a) Externo à Instituição - LUCAS REMOALDO TRAMBAIOLLI - Harvard