Keynote Speakers

Prof. Dr.-Ing. Walter Kellermann

Walter Kellermann is a professor for communications at the University of Erlangen-Nuremberg (FAU), Germany, since 1999. He received the Dipl.-Ing. (univ.) degree in Electrical Engineering from FAU, in 1983, and the Dr.-Ing. degree from the TH Darmstadt, Germany, in 1988. From 1989 to 1990, he was a Postdoc at AT&T Bell Laboratories, Murray Hill, NJ. He authored or coauthored 21 book chapters, 300+ refereed papers in journals and conference proceedings, and 70+ patents, and is a co-recipient of ten best paper awards. Aside from various roles in the IEEE Signal Processing Society, he served as an Associate Editor and Guest Editor to various IEEE and EURASIP journals. His current research interests include speech signal processing, array signal processing, machine learning, and its applications to acoustic human–machine interfaces and autonomous systems. He was awarded the Julius von Haast Fellowship by the Royal Society of New Zealand in 2012 and the Group Technical Achievement Award of the European Association for Signal Processing (EURASIP) in 2015. He is an IEEE Fellow since 2008 and was elevated to EURASIP Fellow in 2021.

Title: Acoustic Signals – Challenging Signal Processing in an Era of Machine Learning

Abstract: Not only since asking for real-time adaptive filtering with more than a thousand parameters forty years ago, acoustic signals continuously challenged the rapidly growing body of signal processing algorithms for seemingly simple tasks, such as speech enhancement, echo control, or source localization, e.g., for hands-free communication in teleconferencing or mobile telephony. Riding the waves of progress in signal processing theory and hardware, increasingly complex tasks were tackled yielding solutions that were not only beneficial in the short term, but also fueled even more research. In the light of some past developments, we will try to identify fundamental trends and present examples for the state of the art in the area of acoustic scene analysis and signal enhancement, characterized by a wave of algorithms exploiting recent advances in machine learning, and often intertwined with traditional model-based methods. Arguing for the continued relevance of this field, we will discuss hitherto unresolved scenarios and unexplored avenues for future acoustic signal processing possibly reaching beyond our current era of machine learning.

Prof. Dr. Sabine Süsstrunk

Prof. Dr. Sabine Süsstrunk leads the Image and Visual Representation Lab in the School of Computer and Communication Sciences (IC) at EPFL since 1999. Since 2015, she is also Director of the Digital Humanities Institute (DHI), College of Humanities (CdH). Her main research areas are in computational photography, computational imaging, color image processing and computer vision, machine learning, and computational image quality and aesthetics. Sabine has authored and co-authored over 150 publications, of which 7 have received best paper/demo awards, and holds over 10 patents.

Sabine served as chair and/or committee member in many international conferences on image processing, computer vision, and image systems engineering. She is Founding Member and Member of the Board (President 2014-2018) of the EPFL-WISH (Women in Science and Humanities) Foundation, Member of the Foundation Council of the SNSF (Swiss National Science Foundation), Member of the Board of the SRG SSR (Swiss Radio and Television), and Member of the Board of Largo Films. She received the IS&T/SPIE 2013 Electronic Imaging Scientist of the Year Award for her contributions to color imaging, computational photography, and image quality, and the 2018 IS&T Raymond C. Bowman Award for dedication in preparing the next generation of imaging scientists. Sabine is a Fellow of IEEE and IS&T.

Title: Opponency revisited

Abstract: According to the efficient coding hypothesis, the goal of the visual system should be to encode the information presented to the retina with as little redundancy as possible. From a signal processing point of view, the first step in removing redundancy is de-correlation, which removes the second order dependencies in the signal. This principle was explored in the context of trichromatic vision by Buchsbaum and Gottschalk (1) and later Ruderman et al. (2) who found that linear de-correlation of the LMS cone responses matches the opponent color coding in the human visual system.

And yet, there is comparatively little research in image processing and computer vision that explicitly model and incorporate color opponency into solving imaging tasks. A common perception is that “colors” are redundant and/or too correlated to be of any interest, or that they are too complex to deal with. Within deep learning frameworks, color features are rarely considered.

In this talk, I will illustrate with several simple examples, such as saliency and super-pixels, that considering opponent colors can significantly improve image processing and computer vision tasks not only in image enhancement but also image segmentation, image ranking, etc. We have in addition extended the concept of “color opponency” to include near-infrared. And we found that de-correlation concepts also apply to deep learning models in rather interesting ways.

(1) http://rspb.royalsocietypublishing.org/content/220/1218/89.short
(2) http://www.opticsinfobase.org/josaa/abstract.cfm?&uri=josaa-15-8-2036

Rémi Gribonval

Rémi Gribonval is a Research Director (Directeur de Recherche) with Inria in the DANTE research group of Laboratoire de l’Informatique du Parallélisme at École Normale Supérieure de Lyon, and the former scientific leader of the PANAMA research group on sparse audio processing at IRISA, Rennes, France. In 2011, he was awarded the Blaise Pascal Award of the GAMNI-SMAI by the French Academy of Sciences, and a starting investigator grant from the European Research Council in 2011. He is an IEEE fellow and a EURASIP Fellow. He founded the series of international workshops SPARS on Signal Processing with Adaptive/Sparse Representations. Since 2002 he has been the coordinator of several national, bilateral and European research projects. He has been a member of the IEEE SPTM Technical Committee and of the SPARS steering committee. He is currently a Senior Area Editor for the IEEE Transactions on Signal Processing.

Rémi Gribonval was a student at Ecole Normale Supérieure, Paris from 1993 to 1997. He received the Ph. D. degree in applied mathematics from the University of Paris-IX Dauphine, Paris, France, in 1999, and his Habilitation à Diriger des Recherches in applied mathematics from the University of Rennes I, Rennes, France, in 2007.

Title: Sparsity, a swiss-knife from inverse problems to deep learning ?

Abstract: Promoting sparse connections in neural networks is natural to control their computational complexity. Given the thoroughly documented role in inverse problems and variable selection, sparsity also has the potentiel to give rise to learning mechanisms endowed with certain interpretability guarantees. Through an overview of recent explorations around this theme, we will compare and contrast classical sparse regularization for inverse problems with multilayer sparse regularization. During our journey we will highlight the potential of an invariant path-embedding of the parameters of a deep network, both to learn the parameters and to analyze their identifiability from the function implemented by the network. We will also observe an unexpected decoupling between the absence of spurious local valleys/minima in the optimization landscape of shallow sparse linear network learning and the tractability of the problem. In the process, we will be remembered that there is life beyond gradient descent, as illustrated by an algorithm that brings speedups of up to two orders of magnitude when learning certain fast transforms via multilayer sparse factorization.

Antonio Ortega

Antonio Ortega received his undergraduate and doctoral degrees from the Universidad Politécnica de Madrid, Madrid, Spain and Columbia University, New York, NY, respectively. In 1994 he joined the Electrical Engineering department at the University of Southern California (USC), where he is currently a Professor and has served as Associate Chair. He is a Fellow of the IEEE and EURASIP, and a member of ACM and APSIPA. He is the Editor-in-Chief of the IEEE Transactions of Signal and Information Processing over Networks and recently served as a member of the Board of Governors of the IEEE Signal Processing Society. He has received several paper awards, including the 2016 Signal Processing Magazine award. His recent research work is focusing on graph signal processing, machine learning, multimedia compression and wireless sensor networks. Over 40 PhD students have completed their PhD thesis under his supervision and his work has led to over 400 publications in international conferences and journals, as well as several patents. He is the author of an upcoming book, “Introduction to Graph Signal Processing”, to be published by Cambridge University Press in 2021.

Title: Applications of Graph Signal Processing: Deep Learning and Point Clouds

Abstract: In the last few years, a growing body of work has been developed to extend and complement well known concepts in spectral graph theory, leading to the emergence of Graph Signal Processing (GSP) as a broad research field. GSP methods consider the data attached to the vertices as a “graph-signal” and seek to create new techniques (filtering, sampling, interpolation), similar to those commonly used in conventional signal processing (for audio, images or video), so that they can be applied to these graph signals. In this talk, we first introduce some basic graph signal processing concepts, providing a brief overview of wavelet graph filter banks, graph signal sampling and graph topology learning. We illustrate the basic concepts with example applications in point cloud processing and deep learning.