affiliation not provided to SSRN
Speech enhancement, Cross-domain feature, self-supervised learning, Knowledge Distillation, Homoscedastic uncertainty