Speaker Verification

Dependencies

The following packages is required to run the scripts:

• pytorch
• pytorch-lightning
• librosa

Usage

• Copy the TIMIT dataset to the corresponding folder, like this:

TIMIT
├── TEST
│   ├── DR1
│   ├── DR2
│   ├── ...
│   └── DR8
└── TRAIN
    ├── DR1
    ├── DR2
    ├── ...
    └── DR8

• Extract MFCC features:

python -m src.preprocess

• Train the neural network:

python -m src.embedder

• If you need tune the hyperparameters, edit src/config.py