University of West Attica
Συντονισμένη κίνηση πολλαπλών βραχιόνων με χρήση διεπαφής υπολογιστή-εγκεφάλου
Abstract
dc:descriptionThis thesis addresses the design and implementation of a Brain-Computer Interface (BCI) system for synchronized control of multiple robotic arms through electroencephalography (EEG) signals. The research aims to develop a Python-based, high-performance BCI framework capable of real-time application in robotic arm coordination. The work is structured into two primary sections. In the initial section, a feature-based methodology is utilized, encompassing stages such as data collection, preprocessing, feature extraction, and model training with classification algorithms like Support Vector Machine (SVM), Decision Tree, Xboost and Random Forest. The second section introduces a Multivariate Time Series approach to improve model precision and robustness. This method includes data preprocessing (e.g., normalization), modeling, analysis, and deployment readiness. The architecture integrates convolutional filters with a long short-term memory (LSTM) network, where the convolutional layers extract high-level features, and the LSTM captures long-term dependencies to overcome vanishing gradient issues. The model achieves an accuracy of 98%, making it suitable for real-time applications. This system could be widely applicable in fields requiring accurate, real-time BCI-based action prediction. This approach thus presents a viable solution for EEG-driven action prediction, with scope for further development and optimization. The thesis further performs a comparative analysis of deep learning architectures, including convolutional and bidirectional LSTM layers, recurrent neural networks (RNNs), multilayer perceptrons (MLPs), and Transformer models, focusing on real-time servomotor control with EEG signals to simulate the robotic arms. Using OpenBCI Community Dataset, a deep neural network with convolutional and bidirectional LSTM layers achieved an accuracy of 98% via k-fold cross-validation. Comparative assessments showed that the CNN-LSTM architecture excelled due to its effective capture of both spatial and temporal EEG features. Implemented in Python, this model communicated with an Arduino board for real-time servomotor control, with classification reports and confusion matrices confirming seamless performance. The Arduino system, programmed for serial communication with Python, executed precise motor control based on EEG predictions, demonstrating the potential of BCI in human-robot interaction and assistive technology. The study also identifies critical security challenges in BCI technology. BCIs, while promising in facilitating brain-device interactions, are susceptible to cybersecurity threats, notably Bluetooth vulnerabilities. This research examines Bluetooth-based risks in BCI systems and proposes countermeasures. Attacks like Bluebugging, Bluejacking, Bluesnarfing, BlueBorne, Location Tracking, Man-in-the-Middle, KNOB, BLESA, and Reflection Attack are analyzed for their mechanics and impact on commercial BCI systems, with specific mitigation strategies recommended to secure BCI applications. The innovation of this study lies in its comprehensive comparative analysis and the application of advanced models under real-world conditions, an aspect that has not been sufficiently explored in previous works. The use of various neural network architectures and their evaluation on real data provides significant insights into the development and enhancement of BCI technology. This framework facilitates a better understanding of the challenges and opportunities arising from the use of this technology and lays the foundation for future innovations in the field of human-machine interaction. Overall, the study demonstrates that the integration of deep learning techniques, such as CNN-LSTM, into BCI systems provides a reliable and efficient solution for the prediction and control of robotic systems, with significant potential for application in real-world scenarios. Simultaneously, the implementation of security mechanisms during the development of these systems is essential to ensure their smooth and safe operation. This doctoral research contributes to science through the development of an innovative model that approaches consumer-oriented brain-computer interface (BCI) systems, which have primarily been designed for entertainment purposes. Despite lacking the high medical precision, characteristic of specialized medical BCI systems, these consumer-grade products offer significant advantages, such as affordability and widespread availability to the public. The research examines the potential of repurposing these systems to support individuals with mobility impairments arising either from injuries or neurological disorders, such as myasthenia. Utilizing affordable BCI systems opens new horizons for patient rehabilitation and independence, making this technology accessible to a broader audience and reducing the financial burden associated with traditional medical solutions. Additionally, the research emphasizes the importance of interdisciplinary collaboration among neuroscience, engineering, and computer science to achieve optimal integration of BCI systems into users' everyday lives. This approach may potentially lead to future innovative solutions, enabling individuals with motor impairments to regain a degree of autonomy and actively participate in society.
Degree
thesis:*- Grantor dc:publisher
- University of West Attica
- Year dc:date
- 2025
Author and committee
dc:creator, dc:contributor.*- Authors dc:creator
-
- Angelakis, Dimitrios
- Αγγελάκης, Δημήτριος
Subjects
dc:subject × 23- Διεπαφή εγκεφάλου-υπολογιστή
- Μηχανική μάθηση
- Κυβερνοασφάλεια
- Τεχνητή νοημοσύνη
- Μηχανή διανυσμάτων στήριξης
- Συνελικτικό νευρωνικό δίκτυο
- Μακρά βραχυπρόθεσμη μνήμη
- Ταξινομητής Τυχαίων Δασών
- Επαναλαμβανόμενα νευρωνικά δίκτυα
- Brain computer interfaces
- Μachine learning
- Artificial intelligence
- Cyber security
- Support vector machines
- Convolution neural networks
- Long short term memory
- Random forests
- Recurrent neural networks
- Βιοϊατρική Μηχανική
- Επιστήμες Μηχανικού και Τεχνολογία
- Medical Engineering
- Engineering and Technology
- Biomedical Engineering
Rights
- Language dc:language
- gre
Identifiers
dc:identifier.*- Identifier
- 10.12681/eadd/58833