Objectives: Stochastic processes appear in many signal processing applications. This course studies continuous time stochastic processes, discrete time random sequences including Markov chains and ARMA sequences.

Program: Markov Chains, Queueing files, ARMA time series

Bibliography : A. Papoulis and S. U. Pillai, Probability, Random Variables and Stochastic Processes , McGraw Hill Higher Education, 4th edition, January 1, 2002.

Professor: Jean-Yves Tourneret

Hours: 9 lectures, 2h written exam

ECTS :2

Objectives: Understand mathematical methods for the derivation of parameter estimators and statistical tests.

Program: Decision theory : Neyman-Pearson Test, Bayes Test, Composite hypotheses - Estimation theory: Qualities of an estimator, Maximum likelihood method, Bayes decision theory.

 Professor: Marco Lops

Hours: lectures: 7 lectures, 2 Exercises, 2h written exam

ECTS : 1.

Objectives: Appropriate signal representations allow to improve the performance of decision (estimation, detection, classification) or compression. This course presents a wide diversity of representations for deterministic and random signals.

Program: Signal class definition - Classical representations (correlation function, spectral density) - Bases of functions (Fourier, Haar, Hadamard, ...) – Time-frequency representations (Short term Fourier Transform, energy distributions – Cohen class : Wigner-Ville,...) – Time-scale representations (continuous wavelet transform, orthogonal and bi-orthogonal discrete wavelet transforms, frames, multi-resolution analysis).

Bibliography: L. Cohen, Time Frequency Analysis, Prentice Hall, 1995.

Professor: Marie Chabert

Hours: 7 lectures, 3 exercises, 2 practical sessions, 2h written exam.

Objectives: Understand advanced filter synthesis methods : optimal filters, QMF filters.

Program: Digital filters : optimization, non-standard structures, implementation – Fast transforms : unitary transforms, applications – 2D processing.

Bibliography: P.P.Vaidyananthan, Multirate Systems and FilterBanks, Prentice Hall, 1995.

F.Castanié, Traitement Numérique du Signal : 2 Méthodes Avancées, polycopié N7, 1997.

Professor: Nicolas Dobigeon

Hours: 7 lectures, 2 exercises, 2 practical sessions, 2h written exam

Objectives: Understand the main principles of classification and pattern recognition, in supervised (a learning set is available) or unsupervised scenarios (no learning set is available)

Program : Feature selection and extraction : Principal component analysis, Fisher criterion, discriminant analysis - Bayesian classification - Linear discriminant functions, neural networks, support vector machines

Bibliography : R. O. Duda, P. E. Hart and D. G. Stork, Pattern Classification, John Wiley & Sons, NY 2001.

Professor: Jean-Yves Tourneret

Hours: 7 lectures, 2h written exam

Objectives: Spectral analysis based on parametric modeling: different models, parameter estimation and application of these modeling.

Program: Parametric modeling: stationary ARMA models, ARIMA and evolutive models - Prony modeling: deterministic and stationary models – Geometrical approach: singular value decomposition, signal and noise subspaces, Pisarenko model, MUSIC – Parametric estimation: maximum likelihood, mean squared error minimization - Other models: mean squared approach, singular value filtering.

Bibliography: S.Kay, Modern Spectral Analysis, Prentice Hall, 1988 (one given per student)

F.Castanié, Analyse Spectrale, Hermès (given for the whole scholar period, one per student)

Professor: Olivier Besson

Hours: 9 lectures, 2h written exam.

Objectives: Arrays of sensors, which consist of spatially distributed antennas, enable one to achieve better signal detection, filtering and localization, by exploiting the spatial dimension and properly combining the signals received on each antenna. This is typically the case of radar applications where target detection and localization can be enhanced or in communications where space diversity can be utilized. This course presents the various means to recover a signal of interest using an array of sensors.

Program: Modelling of signals received on an array of sensors, spatial filtering and beamforming, adaptive arrays, direction of arrival estimation.

Bibliography: H.L. Van Trees, Optimum Array Processing, John Wiley, 2002.

D.G. Manolakis, V. Ingle, S. Kogon, Statistical and Adaptative Signal Processing, Mc Graw Hill, 2000.

Professor: Olivier Besson

Hours: 10 lectures, Matlab project and/or 2h written.

Objectives: Present the main adaptive algorithm and their application in signal processing and telecommunications.

Program: Optimal Wiener filter - Least Mean Square algorithm. – Recursive Least Square Algorithm. Application to equalization, noise reduction, adaptive line enhancement, channel identification…

Bibliography: S.Haykin, Adaptive filter theory, Prentice Hall., 1986.

Professor: Marie Chabert

Hours: 5 lectures, 2 practical sessions, 1h written exam.

Objectives: To study and understand the different kinds of data compression methods.

Program: Introduction, definition of comparison criteria between different compression technics (compression rate, distorsion rate) - Lossless data compression : dictionnary based methods (Ziv-Lempel), arithmetical coding - Scalar quantization : PCM, log-PCM - Predictive coding : DPCM, ADPCM, Delta modulation - Transform coding : DCT, Karhunen-Loève, Hadamard, wavelet transforms, subband coding - Vector quantization : main ideas, optimal quantization.

Professor: Corinne Mailhes

Hours: 9 lecture, 2h written exam

ECTS : 1.5.

Objectives: To give fundamentals in acoustic signal analysis and theoretical models used in speech recognition and synthesis. Learning problems and adaptive algorithms are detailed. Within the field of music processing, analysis and automatic identification methods are given. Still lots of problems are not solved: polyphonic music, instantaneous speech, natural speech, data information search in audio documents…this lecture gives some solutions to begin with these problems.

Program: Fundamentals in speech and music analysis – Fourier Transform, Cepstral analysis, Coding transform – Statistical modelling for automatic speech and music recognition, mixed Gaussians, Hidden Markov Models, Neural Networks – Learning: EM algorithms and its adaptive forms, Evaluation mean: corpus, confident measure… - Speech synthesis – Applications.

Bibliography: Fundamentals of Speech Recognition, L.Rabiner, B.H. Juang, Prentice Hall Signal Processing Series, 1993.

Professor: Régine André-Obrecht

Hours: 7 lecture, 3 practical sessions, 2h written exam.

ECTS : 1.5.

Objectives: Overview of the different kinds of algorithms in speech and audio coding

P rogram: Speech signal properties - Waveform coding - Vocoders : LPC, RELP, MBE - Analysis by synthesis coding : MPE, CELP - Audio Coding : MPEG 1, 2 and 4

Bibliography: A.Spanias, Speech Coding, a tutorial review, Proceedings of the IEEE, Oct. 1994
Davis Pan,A tutorial on MPEG/Audio compression ,  Trans on IEEE Multimedia, 1995, pp. 60-74 (dispo web)
MPEG Audio and Video Coding, IEEE Signal Processing Magazine, Sept 97, vol 14, n5.

Professor: Corinne Mailhes

Hours: 7 lectures, 4h oral exam

ECTS : 1.

 
Objectives:  Application of the advanced signal processing courses on real data.

Program: Bibliographical study and matlab programming.

 Professor: Nicolas Dobigeon.

Hours: 15 practical sessions, 2h exam

Objectives: How to implement real time algorithms of signal processing & image processing
Program: DSP Architecture : Harvard Architecture, MAC, DAG, Program sequencer Pipeline Cache Memory, DMA. New embedded peripheral. Power management. Code secure.  The developed project will be for example dedicated to movement detection in images acquired through a camera.
 Bibliography: Real-Time Digital Signal Processing, : Implementations, Application ...

Professor: Richard Salvetat
Hours: 6h lecture, 32h practical sessions on Blackfin AD_BF533

Objectives: These courses deal with basic and some advanced key concepts in medical imaging, through signals, images and systems.

Program: - Introduction to Medical Imaging and Image quality assessment - Physics of Nuclear Medicine, Emission Computed Tomography (PET and SPECT) imaging and relevant Image processing - Physics of radiology, Projection radiology, Computed Tomography (CT) and relevant Image processing - Reconstruction technique in Tomography -Physics of Nuclear Resonance, Magnetic Resonance Imaging, and relevant Image processing-Physics of Ultrasound, Ultrasound Imaging and Elastography and relevant Image processing-Ultrasound Doppler imaging-Multi-modality Imaging and Related Applications : registration, fusion,..-Medical Images standards-Examples of research application in medical imaging.

Professor: Denis Kouamé and Adrian.Basarab

Hours:30h lecture, 2h written exam, 20h tutored project.

 Objectives:  Overview of radar remote sensing imagery technics.
Program: Introduction to RF technics for remote sensing –General principles and examples of radar images – Technology of Synthetic Aperture Radar – Radar radiometry and altimetry  - Radar  Link budget- Radar interferometry and polarimery- Radar image processing : filtering, segmentation, estimation. Geometry of radar images.

 Professor: Jean-Claude Souyris, Thierry Amiot, Jordi Inglada, Nadine Pourthié.
Hours: 7 lectures , 2h exam

 Objectives:  Application of the  biomedical image processing course on real data.
Program: Bibliographical study and matlab programming.

 Professor: Denis Kouame
Hours: 10 practical sessions, 2h exam

 Objectives:  Application of the  remote sensing radar image course on real data.
Program: Bibliographical study and matlab programming.

 Professor: Marie Chabert.
Hours: 10 practical sessions, 2h exam