| Titre : |
Hybrid deep learning based speech signal separation |
| Type de document : |
document électronique |
| Auteurs : |
Mohamed Yacine Bouaouni, Auteur ; Rayane Ait-Ali-Yahia, Auteur ; Mohamed Arezki Adel Belouchrani, Directeur de thèse |
| Editeur : |
[S.l.] : [s.n.] |
| Année de publication : |
2021 |
| Importance : |
1 fichier PDF (6.64 Mo) |
| Présentation : |
ill. |
| Note générale : |
Mode d'accès : accès au texte intégral par intranet.
Mémoire de Projet de Fin d’Études : Électronique : Alger, École Nationale Polytechnique : 2021
Bibliogr. f. 136 - 141 |
| Langues : |
Anglais (eng) |
| Mots-clés : |
Deep Learning NMF DNN Autoencoders Unfolding algorithm |
| Index. décimale : |
PN00521 |
| Résumé : |
Audio source separation is a challenging problem which consists of identifying the different sources present in a mixed signal, either by using traditional model based methods or using deep learning algorithms. In this work, we propose two different paradigms for combining model based methods (nonnegative matrix factorization) with neural networks to take advantage of both. The first approach fuses the NMF and a deep neural network (DNN) in a two sequential stages stack, where the DNN enhances the separation of the signals by updating the spectrograms/gains that were estimated using the NMF.
Two architectures based on autoencoders are presented in this thesis, that handle two different kind of input data. The second approach is based on the deep unfolding paradigm. It consists of unrolling the optimization algorithm of the model based method into layers of a deep network, and train it using deep learning techniques. |
Hybrid deep learning based speech signal separation [document électronique] / Mohamed Yacine Bouaouni, Auteur ; Rayane Ait-Ali-Yahia, Auteur ; Mohamed Arezki Adel Belouchrani, Directeur de thèse . - [S.l.] : [s.n.], 2021 . - 1 fichier PDF (6.64 Mo) : ill. Mode d'accès : accès au texte intégral par intranet.
Mémoire de Projet de Fin d’Études : Électronique : Alger, École Nationale Polytechnique : 2021
Bibliogr. f. 136 - 141 Langues : Anglais ( eng)
| Mots-clés : |
Deep Learning NMF DNN Autoencoders Unfolding algorithm |
| Index. décimale : |
PN00521 |
| Résumé : |
Audio source separation is a challenging problem which consists of identifying the different sources present in a mixed signal, either by using traditional model based methods or using deep learning algorithms. In this work, we propose two different paradigms for combining model based methods (nonnegative matrix factorization) with neural networks to take advantage of both. The first approach fuses the NMF and a deep neural network (DNN) in a two sequential stages stack, where the DNN enhances the separation of the signals by updating the spectrograms/gains that were estimated using the NMF.
Two architectures based on autoencoders are presented in this thesis, that handle two different kind of input data. The second approach is based on the deep unfolding paradigm. It consists of unrolling the optimization algorithm of the model based method into layers of a deep network, and train it using deep learning techniques. |
|
Ajouter le résultat dans votre panier
Visionner les documents numériques
Faire une suggestion Affiner la rechercheHybrid deep learning based speech signal separation (2021)
