Matches in SemOpenAlex for { <https://semopenalex.org/work/W2016084804> ?p ?o ?g. }
Showing items 1 to 89 of
89
with 100 items per page.
- W2016084804 abstract "Recently, the hybrid model combining deep neural network (DNN) with context-dependent HMMs has achieved some dramatic gains over the conventional GMM/HMM method in many speech recognition tasks. In this paper, we study how to compete with the state-of-the-art DNN/HMM method under the traditional GMM/HMM framework. Instead of using DNN as acoustic model, we use DNN as a front-end bottleneck (BN) feature extraction method to decorrelate long feature vectors concatenated from several consecutive speech frames. More importantly, we have proposed two novel incoherent training methods to explicitly de-correlate BN features in learning of DNN. The first method relies on minimizing coherence of weight matrices in DNN while the second one attempts to minimize correlation coefficients of BN features calculated in each mini-batch data in DNN training. Experimental results on a 70-hr Mandarin transcription task and the 309-hr Switchboard task have shown that the traditional GMM/HMMs using BN features can yield comparable performance as DNN/HMM. The proposed incoherent training can produce 2-3% additional gain over the baseline BN features. At last, the discriminatively trained GMM/HMMs using incoherently trained BN features have consistently surpassed the state-of-the-art DNN/HMMs in all evaluated tasks." @default.
- W2016084804 created "2016-06-24" @default.
- W2016084804 creator A5001894801 @default.
- W2016084804 creator A5008456610 @default.
- W2016084804 creator A5024444797 @default.
- W2016084804 creator A5057227915 @default.
- W2016084804 date "2013-05-01" @default.
- W2016084804 modified "2023-09-30" @default.
- W2016084804 title "Incoherent training of deep neural networks to de-correlate bottleneck features for speech recognition" @default.
- W2016084804 cites W11553578 @default.
- W2016084804 cites W1993882792 @default.
- W2016084804 cites W2012897754 @default.
- W2016084804 cites W2020144989 @default.
- W2016084804 cites W2046899276 @default.
- W2016084804 cites W2063689849 @default.
- W2016084804 cites W2100495367 @default.
- W2016084804 cites W2105099419 @default.
- W2016084804 cites W2123237149 @default.
- W2016084804 cites W2147768505 @default.
- W2016084804 cites W2160306971 @default.
- W2016084804 cites W2394932179 @default.
- W2016084804 cites W3100231466 @default.
- W2016084804 doi "https://doi.org/10.1109/icassp.2013.6639015" @default.
- W2016084804 hasPublicationYear "2013" @default.
- W2016084804 type Work @default.
- W2016084804 sameAs 2016084804 @default.
- W2016084804 citedByCount "35" @default.
- W2016084804 countsByYear W20160848042014 @default.
- W2016084804 countsByYear W20160848042015 @default.
- W2016084804 countsByYear W20160848042016 @default.
- W2016084804 countsByYear W20160848042017 @default.
- W2016084804 countsByYear W20160848042018 @default.
- W2016084804 countsByYear W20160848042019 @default.
- W2016084804 countsByYear W20160848042020 @default.
- W2016084804 countsByYear W20160848042021 @default.
- W2016084804 countsByYear W20160848042022 @default.
- W2016084804 countsByYear W20160848042023 @default.
- W2016084804 crossrefType "proceedings-article" @default.
- W2016084804 hasAuthorship W2016084804A5001894801 @default.
- W2016084804 hasAuthorship W2016084804A5008456610 @default.
- W2016084804 hasAuthorship W2016084804A5024444797 @default.
- W2016084804 hasAuthorship W2016084804A5057227915 @default.
- W2016084804 hasConcept C138885662 @default.
- W2016084804 hasConcept C149635348 @default.
- W2016084804 hasConcept C151730666 @default.
- W2016084804 hasConcept C153180895 @default.
- W2016084804 hasConcept C154945302 @default.
- W2016084804 hasConcept C23224414 @default.
- W2016084804 hasConcept C2776401178 @default.
- W2016084804 hasConcept C2779343474 @default.
- W2016084804 hasConcept C2780513914 @default.
- W2016084804 hasConcept C28490314 @default.
- W2016084804 hasConcept C41008148 @default.
- W2016084804 hasConcept C41895202 @default.
- W2016084804 hasConcept C50644808 @default.
- W2016084804 hasConcept C52622490 @default.
- W2016084804 hasConcept C86803240 @default.
- W2016084804 hasConceptScore W2016084804C138885662 @default.
- W2016084804 hasConceptScore W2016084804C149635348 @default.
- W2016084804 hasConceptScore W2016084804C151730666 @default.
- W2016084804 hasConceptScore W2016084804C153180895 @default.
- W2016084804 hasConceptScore W2016084804C154945302 @default.
- W2016084804 hasConceptScore W2016084804C23224414 @default.
- W2016084804 hasConceptScore W2016084804C2776401178 @default.
- W2016084804 hasConceptScore W2016084804C2779343474 @default.
- W2016084804 hasConceptScore W2016084804C2780513914 @default.
- W2016084804 hasConceptScore W2016084804C28490314 @default.
- W2016084804 hasConceptScore W2016084804C41008148 @default.
- W2016084804 hasConceptScore W2016084804C41895202 @default.
- W2016084804 hasConceptScore W2016084804C50644808 @default.
- W2016084804 hasConceptScore W2016084804C52622490 @default.
- W2016084804 hasConceptScore W2016084804C86803240 @default.
- W2016084804 hasLocation W20160848041 @default.
- W2016084804 hasOpenAccess W2016084804 @default.
- W2016084804 hasPrimaryLocation W20160848041 @default.
- W2016084804 hasRelatedWork W1964120219 @default.
- W2016084804 hasRelatedWork W2000165426 @default.
- W2016084804 hasRelatedWork W2114557664 @default.
- W2016084804 hasRelatedWork W2144059113 @default.
- W2016084804 hasRelatedWork W2146076056 @default.
- W2016084804 hasRelatedWork W2385132419 @default.
- W2016084804 hasRelatedWork W2546942002 @default.
- W2016084804 hasRelatedWork W2772780115 @default.
- W2016084804 hasRelatedWork W2811390910 @default.
- W2016084804 hasRelatedWork W3003836766 @default.
- W2016084804 isParatext "false" @default.
- W2016084804 isRetracted "false" @default.
- W2016084804 magId "2016084804" @default.
- W2016084804 workType "article" @default.