PL-EESR: Perceptual Loss Based End-to-End Robust Speaker Representation Extraction | IEEE Conference Publication | IEEE Xplore