Matches in SemOpenAlex for { <https://semopenalex.org/work/W4224951806> ?p ?o ?g. }
Showing items 1 to 82 of
82
with 100 items per page.
- W4224951806 abstract "Recently, supervised speech separation has made great progress. However, limited by the nature of supervised training, most existing separation methods require ground-truth sources and are trained on synthetic datasets. This ground-truth reliance is problematic, because the ground-truth signals are usually unavailable in real conditions. Moreover, in many industry scenarios, the real acoustic characteristics deviate far from the ones in simulated datasets. Therefore, the performance usually degrades significantly when applying the supervised speech separation models to real applications. To address these problems, in this study, we propose a novel separation consistency training, termed SCT, to exploit the real-world unlabeled mixtures for improving cross-domain unsupervised speech separation in an iterative manner, by leveraging upon the complementary information obtained from heterogeneous (structurally distinct but behaviorally complementary) models. SCT follows a framework using two heterogeneous neural networks (HNNs) to produce high confidence pseudo labels of unlabeled real speech mixtures. These labels are then updated, and used to refine the HNNs to produce more reliable consistent separation results for real mixture pseudo-labeling. To maximally utilize the large complementary information between different separation networks, a cross-knowledge adaptation is further proposed. Together with simulated dataset, those real mixtures with high confidence pseudo labels are then used to update the HNN separation models iteratively. In addition, we find that combing the heterogeneous separation outputs by a simple linear fusion can further slightly improve the final system performance." @default.
- W4224951806 created "2022-04-28" @default.
- W4224951806 creator A5056415893 @default.
- W4224951806 creator A5090367911 @default.
- W4224951806 date "2022-01-01" @default.
- W4224951806 modified "2023-09-30" @default.
- W4224951806 title "Heterogeneous Separation Consistency Training for Adaptation of Unsupervised Speech Separation" @default.
- W4224951806 cites W1552314771 @default.
- W4224951806 cites W2127851351 @default.
- W4224951806 cites W2291877678 @default.
- W4224951806 cites W2304609584 @default.
- W4224951806 cites W2460742184 @default.
- W4224951806 cites W2734774145 @default.
- W4224951806 cites W2735663686 @default.
- W4224951806 cites W2905258586 @default.
- W4224951806 cites W2940275453 @default.
- W4224951806 cites W2952218014 @default.
- W4224951806 cites W2962866211 @default.
- W4224951806 cites W2964058413 @default.
- W4224951806 cites W2995166068 @default.
- W4224951806 cites W3015191643 @default.
- W4224951806 cites W3015199127 @default.
- W4224951806 cites W3015654783 @default.
- W4224951806 cites W3015788098 @default.
- W4224951806 cites W3015794161 @default.
- W4224951806 cites W3015912787 @default.
- W4224951806 cites W3016094953 @default.
- W4224951806 cites W3095219613 @default.
- W4224951806 cites W3145364388 @default.
- W4224951806 cites W3163208914 @default.
- W4224951806 cites W3163346546 @default.
- W4224951806 cites W3163652268 @default.
- W4224951806 doi "https://doi.org/10.2139/ssrn.4121081" @default.
- W4224951806 hasPublicationYear "2022" @default.
- W4224951806 type Work @default.
- W4224951806 citedByCount "1" @default.
- W4224951806 countsByYear W42249518062023 @default.
- W4224951806 crossrefType "journal-article" @default.
- W4224951806 hasAuthorship W4224951806A5056415893 @default.
- W4224951806 hasAuthorship W4224951806A5090367911 @default.
- W4224951806 hasBestOaLocation W42249518062 @default.
- W4224951806 hasConcept C119857082 @default.
- W4224951806 hasConcept C139807058 @default.
- W4224951806 hasConcept C153294291 @default.
- W4224951806 hasConcept C154945302 @default.
- W4224951806 hasConcept C15744967 @default.
- W4224951806 hasConcept C169760540 @default.
- W4224951806 hasConcept C205649164 @default.
- W4224951806 hasConcept C2776061190 @default.
- W4224951806 hasConcept C2776436953 @default.
- W4224951806 hasConcept C2777211547 @default.
- W4224951806 hasConcept C28490314 @default.
- W4224951806 hasConcept C41008148 @default.
- W4224951806 hasConceptScore W4224951806C119857082 @default.
- W4224951806 hasConceptScore W4224951806C139807058 @default.
- W4224951806 hasConceptScore W4224951806C153294291 @default.
- W4224951806 hasConceptScore W4224951806C154945302 @default.
- W4224951806 hasConceptScore W4224951806C15744967 @default.
- W4224951806 hasConceptScore W4224951806C169760540 @default.
- W4224951806 hasConceptScore W4224951806C205649164 @default.
- W4224951806 hasConceptScore W4224951806C2776061190 @default.
- W4224951806 hasConceptScore W4224951806C2776436953 @default.
- W4224951806 hasConceptScore W4224951806C2777211547 @default.
- W4224951806 hasConceptScore W4224951806C28490314 @default.
- W4224951806 hasConceptScore W4224951806C41008148 @default.
- W4224951806 hasLocation W42249518061 @default.
- W4224951806 hasLocation W42249518062 @default.
- W4224951806 hasOpenAccess W4224951806 @default.
- W4224951806 hasPrimaryLocation W42249518061 @default.
- W4224951806 hasRelatedWork W1571518467 @default.
- W4224951806 hasRelatedWork W1576801573 @default.
- W4224951806 hasRelatedWork W2001850503 @default.
- W4224951806 hasRelatedWork W2020291234 @default.
- W4224951806 hasRelatedWork W2094520212 @default.
- W4224951806 hasRelatedWork W2326254127 @default.
- W4224951806 hasRelatedWork W2353865532 @default.
- W4224951806 hasRelatedWork W2748952813 @default.
- W4224951806 hasRelatedWork W2899084033 @default.
- W4224951806 hasRelatedWork W87991986 @default.
- W4224951806 isParatext "false" @default.
- W4224951806 isRetracted "false" @default.
- W4224951806 workType "article" @default.