Global conditioning on speaker identification

[basveeling]

And perhaps using a keras embedding layer to learn a representation for speakers?

[malzzz]

I'm actually very interested in this as well. Just to clarify, are you referring to something like FaceNet for voice? Have you done any more research into this area as of late?

[basveeling]

I haven't, but this would be interesting! I don't have any access to NN training hardware right now, but I'd love to see if this works.

[faroit]

would this paper be relevant here?

Also there are a few deep embedded clustering implementations around. Also one in Keras: https://github.com/fferroni/DEC-Keras but I don't know if this one is well tested