| KEYWORD |
Implementation of an on-chip machine learning algorithm for stress and cognitive load analysis: computation cost and execution time analysis
keywords MACHINE LEARNING, PERFORMANCE
Reference persons DANILO DEMARCHI
External reference persons Irene Buraioli (irene.buraioli@polito.it)
Thesis type EXPERIMENTAL
Description Stress and cognitive load are two conditions that alter our mental and emotional processes daily, even influencing decision-making patterns. The alterations are also reflected in our bodies; therefore, several physiological parameters allow us to evaluate these altered states. Among these, we can mention cardiorespiratory signal, body temperature, electrodermal signal, or cerebral oxygenation. The complexity of extracting correlations between these physiological signals and altered stress states and cognitive load, requires machine learning methodologies to evaluate complex patterns. A necessity for any device is the implementation of such on-chip real-time algorithms by evaluating their computational cost, efficiency, performance, and execution time.
The proposed thesis work will be structured as follows:
1) The candidate will be required to perform a state-of-the-art analysis regarding the possibilities on the market for the implementation of real-time machine learning algorithms applied to the study of on-chip physiological signals.
2) Next, the candidate must implement a machine-learning model on the previously selected hardware.
3) Finally, the candidate will have to validate the algorithm's operation on data in a provided dataset, validating the performance in terms of the computational cost required and execution time by verifying the effective real-time of the proposed solution.
This process will enable the candidate to demonstrate the effectiveness of complex real-time physiological parameter analysis algorithms with an on-chip implementation.
Required skills programming in Python/Matlab, programming in C
Deadline 14/12/2024
PROPONI LA TUA CANDIDATURA