Bridging Nonlinear Dynamics and Neural Networks for EEG-Based Brain Signal Analysis.

• Grant number: 25/03426-6
• Support Opportunities: Scholarships in Brazil - Master
• Start date: April 01, 2025
• End date: March 31, 2027
• Field of knowledge: Engineering - Biomedical Engineering
• Principal Investigator: Renato Machado
• Grantee: João Alfredo Santos de Meireles
• Host Institution: Divisão de Engenharia Eletrônica (IEE). Instituto Tecnológico de Aeronáutica (ITA). Ministério da Defesa (Brasil). São José dos Campos , SP, Brazil
• Company: Universidade Estadual de Campinas (UNICAMP). Faculdade de Engenharia Elétrica e de Computação (FEEC)
• Associated research grant: 20/09838-0 - BI0S - Brazilian Institute of Data Science, AP.PCPE

Abstract

Electroencephalography (EEG) is a tool used to investigate brain activity and has played a significant role in the development of applications such as brain-computer interfaces (BCIs), diagnostic methods, and assistive technologies. However, EEG signal analysis presents challenges due to its high dimensionality, inter- and intra-individual variability, and strong presence of noise. Thus, advanced methods are necessary to extract relevant information and enhance the interpretation of brain patterns. This project proposes the integration of nonlinear dynamics techniques and neural networks for EEG analysis, exploring functional connectivity and recurrence quantification as complementary approaches to modeling neural activity.Initially, the physical principles of EEG and the main signal processing pipelines will be studied, covering preprocessing, filtering, and artifact removal techniques. Subsequently, recurrence techniques will be applied to characterize the dynamic patterns of brain signals, extracting metrics that represent their temporal structure. Recurrence analysis will enable the identification of recurrent dynamic states and transitions between different activity patterns, providing a more descriptive representation of the signals. Based on these metrics, functional connectivity will be assessed, investigating how different brain regions interact over time.In the next stage, neural networks will be developed for the classification of motor states based on EEG data, allowing comparisons between conventional approaches and those based on nonlinear dynamics. The integration of these methods aims to enhance feature extraction and improve the generalization capability of deep learning models. Additionally, architectures that combine dynamic information extracted from recurrence with modern learning techniques, such as convolutional neural networks (CNNs) and recurrent neural networks (RNNs), will be explored.Validation will be conducted using previously collected data and publicly available datasets widely used in the scientific community, ensuring the reproducibility of results and facilitating comparisons with existing approaches. The final phase of the project will include the integration of the proposed methods and analysis of the obtained results. The findings will be consolidated in the dissertation and disseminated through scientific publications and conference presentations, contributing to the advancement of neuroengineering and the development of new paradigms for interpreting brain signals.This work seeks to provide new perspectives on the use of dynamic models in EEG analysis, expanding the possibilities for extracting and interpreting neural patterns. The combination of recurrence and deep learning has the potential to offer more robust solutions for applications in computational neuroscience and the diagnosis of neurological disorders.