Do we need to do any noice cancellation or pre-emphasis before feeding to deepspeech

(Hnipun) #1


Do we need to do any noise cancellation or pre-emphasis on the speech signal before feeding to the DeepSpeech model for inference. Or is there such a approach inside the DeepSpeech Inference model.

Thanks in Advance

(Lissyx) #2

We are working on improving the training to make the model deal better in noisy environment, so, as of today, it could only help if you do some noise cancellation before hand.