This repository contains the code for generating the dataset, train the model and use steernet.
Here are some audio samples that demonstrate the separation performance of SteerNet with the geometry of various commercially available microphone arrays. Note that the reverberation and signal-to-noise ratio vary from one sample to the other, which can also impact the separation performances.
ReSpeaker USB: [Reference] [Mixture] [Separated]
ReSpeaker Core: [Reference] [Mixture] [Separated]
Matrix Creator: [Reference] [Mixture] [Separated]
Matrix Voice: [Reference] [Mixture] [Separated]