R-Spin: Efficient Speaker and Noise-invariant Representation Learning with Acoustic Pieces

Open in new window