| KEYWORD |
Area Engineering
Non-invasive multi-modal stress monitoring
keywords ARTIFICIAL INTELLIGENCE, DEEP LEARNING, PHYSIOLOGICAL SIGNALS, STRESS, WEARABLE DEVICES
Reference persons LUIGI BORZI'
Research Groups DAUIN - GR-24 - reSilient coMputer archItectures and LIfE Sci - SMILIES
Thesis type APPLIED, EXPERIMENTAL
Description Stress is a key factor in various physical and mental health conditions, yet current monitoring methods are either invasive or unreliable. Multi-modal sensor data from wearable devices provides a non-invasive approach to continuously monitor stress levels. Deep learning techniques can leverage these data streams to accurately assess and predict stress, offering a scalable solution for early detection and intervention.
The student will analyse public datasets that include signals such as electroencephalography, electrodermal activity, photoplethysmogram, acceleration, and skin temperature from wearable sensors, enabling multi-modal stress detection. Stress labels are included, often based on standardized stress-inducing tasks or questionnaires.
Lightweight and explainable deep learning models, like convolutional, recurrent or hybrid neural networks, will be applied to fuse the multi-modal data. The goal is to develop an interpretable, fast model to accurately assess stress levels based on physiological signals.
The model is expected to accurately classify stress levels, identify the most influential physiological signals, and provide a practical, non-invasive solution for continuous stress monitoring and early intervention.
Required skills Signal processing; Matlab; Python; Artificial intelligence; Deep learning
Deadline 01/03/2025
PROPONI LA TUA CANDIDATURA