These are examples to support our AVSS 2022 paper.

Reference:
@{Improved Separation of Closely-spaced Speakers by Exploiting Auxiliary Direction of Arrival Information within a U-Net Architecture}

In this work, we focused on improving speaker separation in the case where two concurrent talkers are situated very close to each other. We added auxiliary location information to the input of seech separation neural networks. Below , you can find a video explaning the approach and can find more details in our paper. Additionally, we provide some audio examples for both closely spaced (20 degrees apart) and largly spaced (100 degrees apart) sources (navigation on top).

Explanation video!

contact: stijn.kindt@ugent.be