Skip to content

Implementation of a fully convolutional neural network for speech enhancement

Notifications You must be signed in to change notification settings

Paul-Aime/speech_enhancement

Repository files navigation

speech_enhancement

Implémentation inspirée de : 'A Fully Convolutional Neural Network for Speech Enhancement'.

Prérequis

Installer conda et exécuter :

conda env create -f cpu-environment.yml
conda activate torch

ou :

conda env create -f gpu-environment.yml
conda activate torch-gpu

(tested with cuda 10.1)

ensuite on peut directement utiliser :

python train.py

Exemple de résultat pour un RSB d'entrée de 0 dB

Exemple de débruitage de STFT

Datasets

Utilisé

Parole : TIMIT Acoustic-Phonetic Continuous Speech Corpus

Autres

Parole

Bruit

About

Implementation of a fully convolutional neural network for speech enhancement

Resources

Stars

Watchers

Forks

Releases

No releases published

Packages

No packages published

Languages