Speaker Verification Predict

[Kailegh]

Hi, first of all, thank you for your awesome work. I intend to use this for speaker verification, I know a speaker and I need to know if another audio belongs to him or not. I have some questions regarding your repo:

• I had thought of saving the speaker embeddings, that is equivalent to the part of the code where you calculate the centroid right? I could calculate the centroid of all my known speakers and just load it later.

• As I said my intention is to give the network and audio and test it against a known speaker, but I am a bit lost on how could I do that with your code, could you point me in the right direction? I get lost after you calculate the sim_matrix and the FAR FRR and thresh parameters.

• I think that my main question is how do I fix that threshold parameter and how do I get a binary output meaning if two inputs correspond to the same person or not.

Once more thank you for your work!