Our graduation project aims to help to facilitate the communication between the people who suffer from dysarthria with the people around them in the emergencies. our project pipeline consists of three phases. Automatic speech recognition using bi�directional LSTM, Auto correction using Noisy channel model and Text-To-Speech using google API. the data features are extracted using MFCC.
Arabic dataset:
https://github.com/DoniaWaelSamir/Arabic-Dysarthria-Dateset.git
TORGO dataset :
http://www.cs.toronto.edu/~complingweb/data/TORGO/torgo.html
UASpeech dataset :
http://ifp-08.ifp.uiuc.edu/protected/UASpeech/
username: isle password: download
OpenSource Tools: