Audionumérique
Objectives
- Understand the properties of the audio signal (speech and music)
- Know how to process and model the audio signal
Description
- Introduction to the speech signal, description of the production and human perception of speech. Practical exercises.
- Acquisition of the audio signal by the computer
- Parameterization of the speech signal (MFCC, PLP). Practical application in the lab.
- Modeling of the speech signal (HMM, GMM, DNN). Implementation of a keyword recognition application in practical training (DNN).
Bibliography
- Calliope & Fant (1989). La parole et son traitement automatique. Masson, Paris.
- Mariani, « Analyse, synthèse et codage de la parole », Hermès, Lavoisier, juillet 2002
- Haton, Cerisara, Fohr, Laprie, Smaïli, Reconnaissance automatique de la parole : du signal à son interprétation, Dunod, Paris, 2006
- Hinton & co, « Deep Neural Networks for Acoustic Modeling in Speech Recognition: The Shared Views of Four Research Groups », Signal Processing Magazine, IEEE, vol. 29, n°6, pp. 82-97, nov 2012
- Environnement Google colab : https://colab.research.google.com
Pre-requisites
Bayesian modeling