Speaker-Verification-paris8-

Steps to run:

1. add wav folder in data directory
2. ./extract-features.sh
3. ./normalization-features.sh
4. ./train-world.sh

⚠️ The TrainWorld.cfg from Alize requires .ndx files (to create them see below)

Getting .ndx files to train the models

They can be created with the command:

• ./generate-ndx-files.sh

Their content looks like this:

• abcd f1 f2 f3
• abcd: name of the speaker model to create
• f1, f2, etc: name of the feature files you want to use to create the model

Credits

https://alize.univ-avignon.fr
https://www.irisa.fr/metiss/guig/spro/spro-4.0.1/spro_3.html