Rényi Entropy-Based Shrinkage Algorithm for Sparse Time-Frequency Distribution Reconstruction Using Component Alignment Map
Abstract
Time-frequency distributions (TFDs) are powerful tools for analyzing non-stationary signals, addressing the limitations of the conventional Fourier transform. Compressive Sensing (CS) has emerged as an advanced technique in this field, enabling the reconstruction of a signal's time-frequency distribution from a sparse subset of ambiguity function samples. While this CS-based method demonstrates high performance, determining the optimal regularization parameter remains a significant challenge. A recent approach leverages the local Rényi entropy (LRE) to estimate the local number of components in both time and frequency, replacing the conventional regularization parameter. This led to the development of a Rényi-entropy-based shrinkage algorithm, which outperforms traditional algorithms. However, this algorithm's performance is constrained by the limitations of LRE itself. In this talk, I will discuss these challenges and present an improved approach using the component alignment map (CAM). CAM identifies and extracts regions of the TFD with similar components, enabling more accurate estimation of local component counts and simplifying the Rényi-entropy-based algorithm. Experimental results demonstrate the efficacy of this enhancement, offering improved reconstruction performance compared to existing methods. Furthermore, this work opens new research avenues, including the integration of machine learning techniques to further refine this CS-based method.
Short Biography
Dr. Vedran Jurdana is a Postdoctoral Researcher at the Department of Automation and Electronics, Faculty of Engineering, University of Rijeka (Croatia). He earned his Ph.D. in Electrical Engineering from the same institution and has published in esteemed international journals and conferences, primarily in signal processing. He actively contributes to scientific projects, including a Croatian Science Foundation-funded study investigating early behavioral markers of developmental alterations in visuospatial processing and visual-motor integration in preterm infants. He is a Reviewer of several prominent international journals and conferences and has been a Technical Program Committee member for the International Conference on Broadband Communications for Next Generation Networks and Multimedia Applications. He is serving as a Guest editor of a Special Issue in Sensors journal. His master’s thesis has been completed in collaboration with Johannes Kepler University Linz (Austria) during his ERASMUS student mobility. Over the past several years, he gained further research and teaching experience through several visits to universities in Austria, Slovenia, Hungary, and Finland. His research interests encompass digital signal processing, time-frequency analysis, compressive sensing, information theory, machine learning, assistive technology, and EEG data processing.
