Matches in SemOpenAlex for { <https://semopenalex.org/work/W3097573669> ?p ?o ?g. }
- W3097573669 abstract "Environmental noises and reverberation have a detrimental effect on the performance of automatic speech recognition (ASR) systems. Multi-condition training of neural network-based acoustic models is used to deal with this problem, but it requires many-folds data augmentation, resulting in increased training time. In this paper, we propose utterance-level noise vectors for noise-aware training of acoustic models in hybrid ASR. Our noise vectors are obtained by combining the means of speech frames and silence frames in the utterance, where the speech/silence labels may be obtained from a GMM-HMM model trained for ASR alignments, such that no extra computation is required beyond averaging of feature vectors. We show through experiments on AMI and Aurora-4 that this simple adaptation technique can result in 6-7% relative WER improvement. We implement several embedding-based adaptation baselines proposed in literature, and show that our method outperforms them on both the datasets. Finally, we extend our method to the online ASR setting by using frame-level maximum likelihood for the mean estimation." @default.
- W3097573669 created "2020-11-09" @default.
- W3097573669 creator A5004777817 @default.
- W3097573669 creator A5010122901 @default.
- W3097573669 creator A5057010207 @default.
- W3097573669 creator A5084286453 @default.
- W3097573669 date "2020-11-04" @default.
- W3097573669 modified "2023-09-26" @default.
- W3097573669 title "Frustratingly Easy Noise-aware Training of Acoustic Models." @default.
- W3097573669 cites W1495679096 @default.
- W3097573669 cites W1524333225 @default.
- W3097573669 cites W1631260214 @default.
- W3097573669 cites W16967297 @default.
- W3097573669 cites W1992475611 @default.
- W3097573669 cites W2047769394 @default.
- W3097573669 cites W2062164080 @default.
- W3097573669 cites W2070707809 @default.
- W3097573669 cites W2079623482 @default.
- W3097573669 cites W2086139506 @default.
- W3097573669 cites W2125336414 @default.
- W3097573669 cites W2128653836 @default.
- W3097573669 cites W2146871184 @default.
- W3097573669 cites W2150769028 @default.
- W3097573669 cites W2237224976 @default.
- W3097573669 cites W2242685705 @default.
- W3097573669 cites W2289394825 @default.
- W3097573669 cites W2400339399 @default.
- W3097573669 cites W2514741789 @default.
- W3097573669 cites W2618099328 @default.
- W3097573669 cites W2620717980 @default.
- W3097573669 cites W2747524273 @default.
- W3097573669 cites W2888867175 @default.
- W3097573669 cites W3007327651 @default.
- W3097573669 cites W3094040572 @default.
- W3097573669 cites W45755234 @default.
- W3097573669 cites W47839664 @default.
- W3097573669 cites W97072897 @default.
- W3097573669 hasPublicationYear "2020" @default.
- W3097573669 type Work @default.
- W3097573669 sameAs 3097573669 @default.
- W3097573669 citedByCount "1" @default.
- W3097573669 countsByYear W30975736692021 @default.
- W3097573669 crossrefType "posted-content" @default.
- W3097573669 hasAuthorship W3097573669A5004777817 @default.
- W3097573669 hasAuthorship W3097573669A5010122901 @default.
- W3097573669 hasAuthorship W3097573669A5057010207 @default.
- W3097573669 hasAuthorship W3097573669A5084286453 @default.
- W3097573669 hasConcept C100675267 @default.
- W3097573669 hasConcept C11413529 @default.
- W3097573669 hasConcept C115961682 @default.
- W3097573669 hasConcept C120665830 @default.
- W3097573669 hasConcept C121332964 @default.
- W3097573669 hasConcept C126042441 @default.
- W3097573669 hasConcept C138885662 @default.
- W3097573669 hasConcept C139807058 @default.
- W3097573669 hasConcept C153180895 @default.
- W3097573669 hasConcept C154945302 @default.
- W3097573669 hasConcept C23224414 @default.
- W3097573669 hasConcept C24890656 @default.
- W3097573669 hasConcept C2775852435 @default.
- W3097573669 hasConcept C2776401178 @default.
- W3097573669 hasConcept C28490314 @default.
- W3097573669 hasConcept C41008148 @default.
- W3097573669 hasConcept C41608201 @default.
- W3097573669 hasConcept C41895202 @default.
- W3097573669 hasConcept C45374587 @default.
- W3097573669 hasConcept C50644808 @default.
- W3097573669 hasConcept C76155785 @default.
- W3097573669 hasConcept C95851461 @default.
- W3097573669 hasConcept C99498987 @default.
- W3097573669 hasConceptScore W3097573669C100675267 @default.
- W3097573669 hasConceptScore W3097573669C11413529 @default.
- W3097573669 hasConceptScore W3097573669C115961682 @default.
- W3097573669 hasConceptScore W3097573669C120665830 @default.
- W3097573669 hasConceptScore W3097573669C121332964 @default.
- W3097573669 hasConceptScore W3097573669C126042441 @default.
- W3097573669 hasConceptScore W3097573669C138885662 @default.
- W3097573669 hasConceptScore W3097573669C139807058 @default.
- W3097573669 hasConceptScore W3097573669C153180895 @default.
- W3097573669 hasConceptScore W3097573669C154945302 @default.
- W3097573669 hasConceptScore W3097573669C23224414 @default.
- W3097573669 hasConceptScore W3097573669C24890656 @default.
- W3097573669 hasConceptScore W3097573669C2775852435 @default.
- W3097573669 hasConceptScore W3097573669C2776401178 @default.
- W3097573669 hasConceptScore W3097573669C28490314 @default.
- W3097573669 hasConceptScore W3097573669C41008148 @default.
- W3097573669 hasConceptScore W3097573669C41608201 @default.
- W3097573669 hasConceptScore W3097573669C41895202 @default.
- W3097573669 hasConceptScore W3097573669C45374587 @default.
- W3097573669 hasConceptScore W3097573669C50644808 @default.
- W3097573669 hasConceptScore W3097573669C76155785 @default.
- W3097573669 hasConceptScore W3097573669C95851461 @default.
- W3097573669 hasConceptScore W3097573669C99498987 @default.
- W3097573669 hasOpenAccess W3097573669 @default.
- W3097573669 hasRelatedWork W1504449689 @default.
- W3097573669 hasRelatedWork W2134749286 @default.
- W3097573669 hasRelatedWork W2145334935 @default.
- W3097573669 hasRelatedWork W2145708999 @default.
- W3097573669 hasRelatedWork W2522791656 @default.
- W3097573669 hasRelatedWork W2572023240 @default.