Signal Enhancement Methods

Catalogue des cours de Télécom SudParis

Code

SIC 7001

Niveau

PostGraduate (MSc)

Période

Spring (P2)

Domaine

Signal et Communications

Langue d'enseignement

English

Crédits ECTS

6

Heures programmées / Charge de travail

24 / 96

Responsable(s)

  • BOUDY Jerome

Département

- Electronique et Physique

Equipe pédagogique

  • BALDINGER Jean louis
  • BOUDY Jerome
  • HOUMANI Nesma

Objectif

- To remind basic notions in signal processing, namely the transformation from the analog domain towards the digital domain (sampling/quantization) and frequency-domain analysis methods
- Introduction to some advanced signal enhancement techniques for noise-embedded measurements contexts in order to improve the output sensors signal-to-noise ratio (SNR): instances of some kinds of signals will be presented (speech, biomedical signals, biometric signals…)
- To introduce time-frequency analysis methods often used for the signal enhancement (in transmission) and extraction of its relevant parameters (after A-D conversion) in Pattern Recognition, in order to improve the global system robustness to adverse environments (interferences).

- N.B.: here we mean by « signal enhancement » (cf. title) all the approaches aiming at analysing in a relevant way (extraction) and/or to reduce noise signal before its analysis.

Contenu

A) Characterisation of different signals (recalls): deterministic/stochastic signals, notion of noise signals, Autocorrelation functions and Spectral Energy/Power Densities.
B) Linear Filtering of Signals: convolution, invariant linear filters, filter transfer functions (in Z and frequency domains).
C) Signal acquisition : analog-to-digital conversion, Sampling principle, quantization (SNR ratio).
D) Digital linear filtering
E) Time-frequency analysis : Discrete & Fast Fourier Transforms (DFT & FFT), Wavelets Transforms, Homomorphic Analysis (Cepstrum, LFCC , MFCC, Root-spectral…).
F) Noise reduction : Wiener filter, Spectral Substraction , Adaptive filtering.
G) Multi-cadence filtering : Filterbanks, complexity reduction

Prérequis

- Basic knowledge in Fourier Analysis and in Statistics

Mots-clés

Sampling, quantization, time-frequency analysis, parameters extraction, linear and adaptive filtering, noise reduction, multi-cadence filtering.

Evaluation

1st session = control (C1)
2nd session = control (C2)
Final mark = Sup (C1,C2)

Approches pédagogiques

 

Programme

Master of Science

Fiche mise à jour : 06/01/2017 15:11:59