Training and Research
PhD Programme Courses/classes
This page shows the PhD course's training activities for the academic year 2024/2025. Further activities will be added during the year. Please check regularly for updates!
Modelli dinamici e simulazione di sistemi multibody
Credits: 5
Language: English
Teacher: Iacopo Tamellin
Advanced techniques for acquisition of biomedical images
Credits: 2.5
Language: Ingelese
Teacher: Paolo Farace, Federico Boschi
Theranostics: from materials to devices
Credits: 1
Language: english
Teacher: Nicola Daldosso, Tommaso Del Rosso
Nanomaterials: synthesis, characterization and applications
Credits: 1
Language: English
Teacher: Francesco Enrichi, Tommaso Del Rosso
Brain Computer Interfaces
Credits: 3
Language: Inglese
Teacher: Silvia Francesca Storti
Algorithmic motion planning in robotics
Credits: 1
Language: Italian
Teacher: Paolo Fiorini
Data visualization
Credits: 1
Language: Inglese
Teacher: Andrea Giachetti
Sistemi Ciber-Fisici nell’Industria 4.0: Modellazione, Reti e Intelligenza
Credits: 3
Language: English
Teacher: Enrico Fraccaroli
Modellazione e analisi 3D
Credits: 1
Language: Inglese
Teacher: Andrea Giachetti
Modelli di Intelligenza Artificiale Spiegabile: stato dell'arte, promesse e sfide
Credits: 2.5
Language: Inglese
Teacher: Gloria Menegaz
Foundation of Robotics Autonomy
Credits: 1
Language: Italian
Teacher: Paolo Fiorini
Generative AI
Credits: 1.5
Language: English
Teacher: Francesco Setti
Modeling and Verification of Digital Systems
Credits: 1.5
Language: Italian
Teacher: Franco Fummi, Nicola Bombieri, Graziano Pravadelli
Soft robotics: from nature to engineering
Credits: 1.5
Language: English
Teacher: Francesco Visentin
Techniques and algorithms for biomechanics of movement
Credits: 2.5
Language: English
Teacher: Roberto Di Marco
Brain Computer Interfaces (2024/2025)
Teacher
Referent
Credits
3
Language
Inglese
Class attendance
Free Choice
Location
VERONA
Learning objectives
The aim of this course is to propose an introduction to the basics of Brain Computer Interfaces (BCI) principally based on oscillatory EEG activity from a signal processing point of view. The course will introduce the main data processing methods that allow to decode brain activity in real time and convert it into a control signal for a BCI. In the first part the students will learn the following topics: the BCI model, the main BCI types with relative basic signal processing techniques for feature extraction and classification, the performance of the systems, the limitations of the current paradigms and the broad range BCI applications. The second part will cover practical BCI design and use, with an introduction to real-time processing of EEG recordings. Collaboration among students with different backgrounds will be encouraged through research-oriented practical group projects.
Prerequisites and basic notions
The recommended prerequisites of the course are basic familiarity with signal processing and programming in Matlab.
Program
- Introduction to the BCI model. Motivation for BCI. Its historical context and recent approaches. The BCI technology.
- Applications: in medicine, prevention of risk situations, smart environments, gaming etc.
- Invasive and non-invasive BCIs
- EEG-based control signals: evoked (e.g., SSVEP and P300 speller) vs. self-paced.
- Signal processing (filtering, feature extraction, classification) and the interpretation of the results.
- Kinesthetic motor imagery and introduction to a typical architecture of EEG-based MI-BCI (calibration and usage phases).
- The role of machine learning in BCIs.
- The classification problem and how to access performances.
- Case studies.
Laboratory. The lab involves implementing an MI-BCI interface in Matlab. Students will use EEGlab to create Matlab scripts and work on EEG-BCI data, filtering the data, extracting features like power spectral density, coherence, and correlation in the frequency bands of interest, implementing a classifier to distinguish different imagined movements. Finally, they will interpret the results obtained.
Bibliography
Didactic methods
Lectures will be conducted both in person and via streaming, providing students abroad with the opportunity to attend remotely. Educational material will be available to students enrolled in the course on the Moodle platform. This material includes lecture presentations in PDF format and material related to laboratory activities.
Learning assessment procedures
The exam consists in developing a short project in Matlab for analyzing EEG-BCI data. This task will require students to apply the knowledge gained during the course, facing challenges related to processing and interpreting brain signals.
Assessment
To pass the exam, the students must show that: - they have understood the theoretical and practical concepts of the course; - they are able to use the knowledge acquired during the course to solve the assigned problems related to the processing of EEG-based BCI signals; - they are able to program in MATLAB environment in the context of signal processing.
Criteria for the composition of the final grade
Pass/no pass.
Scheduled Lessons
| When | Classroom | Teacher | topics |
|---|---|---|---|
|
Wednesday 12 March 2025 10:30 - 12:30 Duration: 2:00 AM |
Ca' Vignal 3 - T.06 [06 - T] | Silvia Francesca Storti | Introduction to Brain-Computer Interfaces (BCIs): definition of BCI; how BCIs work; motivation for BCIs; the BCI model; the role of feedback; types of BCIs (active, reactive, and passive). Current neuroimaging-based BCI modalities. The role of machine learning in BCIs. Offline training and online testing. History of BCIs and recent approaches. |
|
Tuesday 18 March 2025 10:30 - 12:30 Duration: 2:00 AM |
Ca' Vignal 3 - Laboratorio Ciberfisico [103 - ] | Silvia Francesca Storti | Applications of BCIs: medical applications (communication, rehabilitation and restoration, detection and diagnosis); prevention of risk situations (passive BCI), smart environments, neuromarketing, educational, gaming, military use. Design and implementation of BCIs. Signal acquisition methods (invasive and non-invasive BCIs). Focus on non-invasive EEG-based BCIs. |
|
Monday 24 March 2025 13:30 - 15:30 Duration: 2:00 AM |
Ca' Vignal 3 - Laboratorio Ciberfisico [103 - ] | Silvia Francesca Storti | EEG-based BCI control signals: slow cortical potentials; evoked potentials (SSVEP, P300 speller), motor-imagery systems based on sensorimotor desynchronization. Kinesthetic motor imagery and introduction to a typical architecture of EEG-based MI-BCI (experimental paradigm, signal acquisition). |
|
Tuesday 01 April 2025 09:30 - 11:30 Duration: 2:00 AM |
Ca' Vignal 3 - Laboratorio Ciberfisico [103 - ] | Silvia Francesca Storti | Study of a typical architecture of EEG-based MI-BCI: signal preprocessing methods for the removal of physiological and extraphysiological artifacts (temporal and spatial filtering), feature extraction methods based on spectral information (calibration phase), and time-frequency methods for online usage phase, event-related potentials (ERS/ERD), feature storage, the classification problem for MI-BCI systems (training data, predictor function, empirical risk, overfitting, and underfitting problems). |
|
Tuesday 08 April 2025 10:30 - 12:30 Duration: 2:00 AM |
Ca' Vignal 3 - Laboratorio Ciberfisico [103 - ] | Silvia Francesca Storti | Laboratory. The laboratory involves implementing a simple offline MI-BCI interface in Matlab following the architecture explained during the lectures. The laboratory is structured into two main parts: preprocessing+feature extraction and classification. Initially, students receive a description of the experimental paradigm of the provided EEG data with an explanation of the key functions for scripting via EEGLAB (Matlab toolbox). Students uses a draft code and are required to implement some crucial processing steps. The features to extract are: power spectral density, coherence, and correlation for the alpha and beta frequency bands. |
|
Tuesday 15 April 2025 10:30 - 12:30 Duration: 2:00 AM |
Ca' Vignal 3 - Laboratorio Ciberfisico [103 - ] | Silvia Francesca Storti | Laboratory. This second part involves training a support vector machine (SVM) classifier to decode imagined movement of single trials and testing the classifier with cross-validation. Explanation is provided on how to assess the classifier's performance and optimize the algorithm's hyperparameters. The main challenges in BCI classification are highlighted: insufficient quantity of data, poor data quality, non-representative training data, irrelevant features requiring dimensionality reduction (channel/feature selection). Model evaluation involves considerations such as overfitting and underfitting the training data. In this part as well, students use a code template and are tasked with implementing crucial processing steps, focusing on the interpretation of the obtained results. As a small project, students are assigned the processing of new EEG-BCI data. |
Sustainable Development Goals - SDGs
This initiative contributes to the achievement of the Sustainable Development Goals of the UN Agenda 2030. More information on sustainability
