fakufaku / torchiva

Blind source separation with independent vector analysis family of algorithm in torch
https://torchiva.readthedocs.io/en/latest/
MIT License
86 stars 5 forks source link

Beamformers #5

Closed Toplica001 closed 10 months ago

Toplica001 commented 10 months ago

Hello, how to use beamformers (which script do i need to run and how?) and are beamformers AI based or not? And one more thing, on one pc i can install this library but on my main pc i can't, when i try to install it says that i dont have module named fcntl which is the module related to linux. Do u have a solution for that as well or not, on one windows pc works on the other doesn't? Thank you in advance.

fakufaku commented 10 months ago

Hi @Toplica001 , sorry there are no example code for the beamformers, but you could try to modify the training code in the examples folder. The beamformers are hybrid with the target/noise statistics being estimated with a NN and the beamformers are computing according to some traditional optimal formulation.

I don't know about the fcntl module. This is not a direct dependency of torchiva (probably a dependency of a dependency) so I cannot help you here. The main dependency is pytorch, so if you can install it, you should be able to use the separation modules I think.

Toplica001 commented 10 months ago

Ok, thank you for your reply.