MelBank

This project objective is to separate the sound of multiple speakers on a single channel.

It enables not only noise-speech separation, but also speech-speech separation.

Demo

Cannot play demo audio in GitHub. If you want to listen to demo audio, look this.

$ git clone <this repo>
$ cd <this repo>

$ pipenv install

You also need to install portaudio.

$ pipenv run record # Recording each sound source to be separated
$ pipenv run build  # Build teacher data

$ pipenv run train

$ pipenv run demo

If you want to know the details of how to use this, run the following command.

$ pipenv run help