From Traditional Signal Processing to Machine-Learning-Based Signal Processing


Lecturer: Izv. prof. dr. sc. Jonatan Lerga, University of Rijeka

Date: April 3, 2024, 14:00

Location: JNU, GAMF Faculty, 4/110 

Signal processing has undergone a remarkable transformation from traditional methods to machine learning-driven approaches. Traditional techniques, rooted in mathematical principles, are limited in the case of large, real-life datasets. Machine learning-based signal processing offers new possibilities for signal analysis by learning from data to detect intricate patterns within signals. Machine learning-based approaches provide improved performance, accuracy, scalability, and robustness compared to traditional methods. This shift has led to significant advancements in various fields. In the presentation, some of our recently developed signal processing methods for adaptive filtering, denoising, compressive sensing, time-frequency representation, instantaneous frequency estimation, and entropy-based useful information extraction will be presented. Next, machine-learning-based methods we developed/applied to medical imaging, medical signals, remote sensing, underwater image analysis, music transcription, seismology, gravitational-wave analysis, applications in hydrology, and sea state estimation will be presented.

prof. dr. sc. Jonatan Lerga is the Head of the Center for Artificial Intelligence and Cybersecurity at University of Rijeka. At the same time he also serves as Vice Dean for Academic Affairs at the Faculty of Engineering of the University of Rijeka.

All interested are welcome! Participation in the lecture is free of charge.