From 1982 to 1989, I was
associate-professor at ENSERG, electrical engineering
department of Institut National Polytechnique de Grenoble.
Since 1989, I am full professor at Université Joseph
Fourier (UJF) in Grenoble which became
University
Grenoble-Alpes in 2016. Although my main academic
duties were in
Polytech’Grenoble
in the
electrical
engineering department (IESE), I also presented
lectures for graduate students preparing PhD (DEA and
Master research) especially in Cognitive Sciences. Since
September 2019, I am emeritus professor at Univ.
Grenoble-Alpes, without regular academic duties.
In addition to my academic charge,
I frequently presented lectures in other French
universities (UFR IMA Grenoble, ENST Bretagne) and
foreigner universities : Politechnico di Torino (Italy),
Universitat Politechnica de Catalunya et Universitat de
Valencia (Spain), Univ. of Campinas (Brazil), Sharif
University of Technology (Iran), etc., mainly for MSc or
PhD students and in relation with my research
activities.
Next paragraphs provide more information concerning my
academic charge during the last years. The lecture
documents and slides (in French) given to my students can
be downloaded in this
directory.
Devoted for (first year of
Master) students in electrical engineering (3i),
this lecture on advance electronics aims to use
theoretical knowledges for the design of a reliable
electronical device, which is a key practical problem for
a future engineer. This lecture mainly tries to
answer to the following question: how design an electronic
system which is as independent as possible of component
parameters and external conditions, like power supply and
temperature. I stopped to teach this lecture in 2007.
Devoted to last year
students (second year of Master in electrical engineering)
of 3i, option Automatic Control, this lecture is focused
on nonlinear servo-controls, addressed using first
harmonic method, and phase space method. The latter method
is particularly interesting since it can be applied for
studying stability of both physical systems and
algorithms. I stopped to teach this lecture in 2000.
This lecture is focused on
mathematical tools for representing and analysing signals,
deterministic as well as random. First, I present to the
students simple applications in signal and image
processing, for showing that they use everyday signal
processing tools. For instance, I speak them about
spreading spectrum communications, IRM imaging, EEG and
ECG, watermarking, noise cancellation and source
separation. After the description of mathematical tools
for both deterministic and random signals, the last
chapter of this lecture presents usual principles of
signal processing, like correlation methods, matched
filtering and Wiener filtering.
This lecture addresses the
design of linear optimal filters, optimal in the sense of
mean square error minimization. For sake of its shortness,
the problem is restricted to discrete time signals and
filters. After explaining principles of Wiener filtering,
I show how we can design practical algorithms,
like LMS and RLS, including convergence study. The last
chapter is focused on Kalman filter, which is illustrated
by an application for 50 Hz power line rejection developed
in the laboratory by one of my PhD students, Dr. R.
Sameni.
Detection,
estimation and information theories
This lecture presents
basics of statistical theories of decision: Bayes
criterion, Neyman-Pearson criterion, Minimax de Bayes, de
Neyman-Pearson, Minimax, and estimation: least mean
square estimatiion, maximum a posteriori, maximum
likelihood, Cramer-Rao bounds. The last part of the
lecture concerns information theory: Shannon entropies and
mutual information, Kraft and McMillan theorems,
Shannon-Fano and Hufmann codes. This lecture (20 hours) is
illustrated with 20 hours of exercices and problems in
signal processing, communications, pattern recognition or
neural networks.
Artificial
neural networks and applications
This lecture is a panorama
of main neural algorithms (MLP, Kohonen self-organizing
maps, Hopfield models, source separation), illustrated by
applications in signal and image processing: image
denoising and filtering, image compression,
classification, pattern recognition, identification,
prediction, estimation, source separation, etc. I
especially insist on differences, advantages and
drawbacks, between neural methods and classical methods
in classification, identification, vectorial
quantization, etc. This lecture was been first funded in
the framework of an European project, in 1990, and
presented in various cities in Europe (Louvain, Lausanne,
Torino, Barcelona, Granada) and then presented to MSc
students in electrical engineering and in cognitive
sciences.