/kaldi-for-dummies

This is the repository for my version of Kaldi for Dummies example.

Primary LanguageShell

Kaldi for dummies

The digits recordings data has been taken from here.

The data was brought to the required format by running the script script.sh which is there in the root directory of the project. It should be run in the directory where the above audio files exist. You may need to change permissions before running the script.sh

Steps to bring the dataset to proper format.

  1. Move N speakers to test directory and remaining to train
  2. Create a file spk2gender to show the mapping of the speakers to their gender.
  3. Run the scripts script_for_utt2spk.sh, script_for_wav.scp.sh, and script_for_text.sh to generate the respective files.
  4. text, wav.scp, and utt2spk would be generated as a result of the above step. Move these files to their respective test or train sub-directories in the data directory. Completing these steps ensures that the data has been properly set up.

For further steps, kaldi-for-dummies can be followed from here.