I used DeepSpeech for speech recognition for a dataset with 10 speakers. The problem is that after I test it with my voice (which is normal because my voice was not in the dataset), but for the speakers that I used for training works. The question is: is it possible to make the system work for any speaker? even if I train only for 10 speakers. I would appreciate if you know some links that could help me.
Try SpecAugment ? Fine tune on recordings of yourself?