Real time diarization about speaker-diarization HOT 6 OPEN

1: About the 1st question:
Sure, the clustering method was designing to eval the label of each clip of a long audio, with consideration for new coming speakers.

In realtime view, you should reimplement the uisrnnModel.predict function, since it took a list of embeddings as input with fixed size and beam search method.

2: Had better use the same language between your training data and evaluating data, since different language has different tempo and feature, unless there is a more robust speaker recognition method.

from speaker-diarization.

hi @giorgionanfa have you tried the live diarization , i tried to do it but i always get the same speaker [0].

from speaker-diarization.

@taylorlu Could you please give more details about how to reimplement the function? What steps should I take to make it works for the realtime diarization? Thank you very much!

from speaker-diarization.

Ok, thanks so much, i will try

from speaker-diarization.

Hi @rohithkodali, i have not tried because i worked on other topics meanwhile. How have you modified the code?

from speaker-diarization.

@rohithkodali have you any update?

from speaker-diarization.