jtkim-kaist/VAD

How to create the dataset

pankaj2701 opened this issue · 2 comments

pankaj2701 commented

I am wondering how to create the training dataset. I have understood the format but don't know how to create one. Do we have to manually annotate the training dataset. Manually annotation would be difficult, is there any utility which can create an approximate training data which can then refined manually.

pankaj2701 commented

Manual annotation of data with music, noise and speech mixed would be difficult.
Can we provide data with noise, speech or music separately. Labelling would be easier in this case.

jtkim-kaist commented

Our provided recorded dataset was manually annotated by two professional engineers.

If you want to construct the training set, do the followings:

  1. run the VAD to the clean speech to get the label.

  2. Add the noise to that clean speech using FanT tool or voice box(Matlab implemented)

  3. Then you can get the labels with noisy speech

If you don't have clean speech manual annotation cannot be avoidable.