Christian JUTTEN / LIG
Christian Jutten received Ph.D. and Doctor es Sciences degrees in signal processing from Grenoble Institute of Technology (GIT), France, in 1981 and 1987, respectively. From 1982, he was an Associate Professor at GIT, before being Full Professor at Univ. Grenoble Alpes, in 1989. Since 80’s, his research interests have been machine learning and source separation, including theory (separability, nonlinear mixtures, sparsity, multimodality) and applications (brain and hyperspectral imaging, chemical sensor array, speech). He is author or coauthor of 105+ papers in international journals, 4 books, 27 keynote plenary talks and 230+ communications in international conferences.
He has been visiting professor at EPFL (Lausanne, Switzerland, 1989), Riken labs (Japan, 1996) and Campinas Univ. (Brazil, 2010). He was director or deputy director of his lab from 1993 to 2010, especially head of the signal processing department (120 people) and deputy director of GIPSA-lab (300 people) from 2007 to 2010. He was a scientific advisor for signal and images processing at the French Ministry of Research (1996–1998) and CNRS (2003–2006 and since 2012).
Christian Jutten was organizer or program chair of many international conferences, especially of the 1st Int. Conf. on Blind Signal Separation and Independent Component Analysis in 1999 (ICA’99). He has been a member of a few IEEE Technical Committees, especially Machine Learning for Signal Processing and Theory and Methods for Signal Processing. He received many awards, e.g. best paper awards of EURASIP (1992) and IEEE GRSS (2012), Medal Blondel (1997) from the French Electrical Engineering society, and one Grand Prix of the French Académie des Sciences (2016). He was elevated as IEEE fellow (2008), EURASIP fellow (2013) and was a Senior Member of Institut Universitaire de France from 2008 to 2018. He is the recipient of a 2012 ERC Advanced Grant for the project Challenges in Extraction and Separation of Sources (CHESS). He is now emeritus professor at Univ. Grenoble Alpes.
Due to technology advances, multimodal recordings are currently very common, e.g. EEG/MEG/MRI for brain imaging, video and sound recordings for scene analysis or hyperspectral and LIDAR recordings for remote sensing. It is usually believed that such multimodal recordings lead to enhanced information and better estimation. However, what does it happen when the relationships between the modalities are not perfectly known? For answering this issue, we will consider a very simple and comprehensible example involving two modalities, each one associated to a Gaussian noisy information channel. We will analytically derive and discuss the - sometimes surprising - effects of the input prior mismatch and of the noise mismatch on mutual information and excess mean square error.
Updated on:Jan. 1, 2021, midnight
Number of views: