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
- Move N speakers to
test
directory and remaining totrain
- Create a file spk2gender to show the mapping of the speakers to their gender.
- Run the scripts
script_for_utt2spk.sh
,script_for_wav.scp.sh
, andscript_for_text.sh
to generate the respective files. text
,wav.scp
, andutt2spk
would be generated as a result of the above step. Move these files to their respective test or train sub-directories in thedata
directory. Completing these steps ensures that the data has been properly set up.
For further steps, kaldi-for-dummies can be followed from here.