How to separate data set for training the universal background model?

I have a data set of 50 speakers, with 20 utterances of each speaker. In order to train a GMMUBM system can I use features from the first 10 utterances of each speaker to train the UBM model, and...

18 October 2016 9,544 3 View