Matches in SemOpenAlex for { <https://semopenalex.org/work/W3011386605> ?p ?o ?g. }
Showing items 1 to 70 of
70
with 100 items per page.
- W3011386605 abstract "This paper proposes a novel approach of DNN-based statistical parametric speech synthesis where non-negative matrix factorization (NMF) is effectively utilized. In statistical parametric speech synthesis, Mel-frequency cepstrum is often employed for acoustic features. However, it represents a spectral envelope as a linear combination of fixed envelope curves (sines and cosines), and the envelope predicted by a DNN-based acoustic model loses its fine structure. On the other hand, in NMF, multiple spectral envelopes (spectrogram) are decomposed into two factors; spectral bases and their activity patterns (activation). Since the obtained bases keep the fine structure of envelopes, the remaining factor, i.e. activation can be employed for acoustic features. Due to its sparseness, the spectral envelope obtained by the predicted activation also keeps fine structure. In this study, activation derived from NMF is utilized for spectral representation, and DNN-based text-to-speech synthesis incorporating NMF is proposed. In addition, this framework can potentially incorporate some applications of NMF, such as bandwidth expansion, voice conversion, or noise reduction. In this study, bandwidth expansion is achieved, and experimental results demonstrate that the proposed method can generate more natural spectral parameters especially in 48 kHz sampling rate, and that 16 kHz-to-48 kHz bandwidth expansion, where natural synthetic speech is produced, is achieved in the proposed framework." @default.
- W3011386605 created "2020-03-23" @default.
- W3011386605 creator A5010841595 @default.
- W3011386605 creator A5041213266 @default.
- W3011386605 creator A5045273229 @default.
- W3011386605 date "2019-11-01" @default.
- W3011386605 modified "2023-10-16" @default.
- W3011386605 title "DNN-based Statistical Parametric Speech Synthesis Incorporating Non-negative Matrix Factorization" @default.
- W3011386605 doi "https://doi.org/10.1109/apsipaasc47483.2019.9023311" @default.
- W3011386605 hasPublicationYear "2019" @default.
- W3011386605 type Work @default.
- W3011386605 sameAs 3011386605 @default.
- W3011386605 citedByCount "0" @default.
- W3011386605 crossrefType "proceedings-article" @default.
- W3011386605 hasAuthorship W3011386605A5010841595 @default.
- W3011386605 hasAuthorship W3011386605A5041213266 @default.
- W3011386605 hasAuthorship W3011386605A5045273229 @default.
- W3011386605 hasConcept C105795698 @default.
- W3011386605 hasConcept C11413529 @default.
- W3011386605 hasConcept C117251300 @default.
- W3011386605 hasConcept C121332964 @default.
- W3011386605 hasConcept C14999030 @default.
- W3011386605 hasConcept C152671427 @default.
- W3011386605 hasConcept C153180895 @default.
- W3011386605 hasConcept C154945302 @default.
- W3011386605 hasConcept C158693339 @default.
- W3011386605 hasConcept C2776257435 @default.
- W3011386605 hasConcept C28490314 @default.
- W3011386605 hasConcept C33923547 @default.
- W3011386605 hasConcept C41008148 @default.
- W3011386605 hasConcept C42355184 @default.
- W3011386605 hasConcept C45273575 @default.
- W3011386605 hasConcept C54926389 @default.
- W3011386605 hasConcept C62520636 @default.
- W3011386605 hasConcept C76155785 @default.
- W3011386605 hasConceptScore W3011386605C105795698 @default.
- W3011386605 hasConceptScore W3011386605C11413529 @default.
- W3011386605 hasConceptScore W3011386605C117251300 @default.
- W3011386605 hasConceptScore W3011386605C121332964 @default.
- W3011386605 hasConceptScore W3011386605C14999030 @default.
- W3011386605 hasConceptScore W3011386605C152671427 @default.
- W3011386605 hasConceptScore W3011386605C153180895 @default.
- W3011386605 hasConceptScore W3011386605C154945302 @default.
- W3011386605 hasConceptScore W3011386605C158693339 @default.
- W3011386605 hasConceptScore W3011386605C2776257435 @default.
- W3011386605 hasConceptScore W3011386605C28490314 @default.
- W3011386605 hasConceptScore W3011386605C33923547 @default.
- W3011386605 hasConceptScore W3011386605C41008148 @default.
- W3011386605 hasConceptScore W3011386605C42355184 @default.
- W3011386605 hasConceptScore W3011386605C45273575 @default.
- W3011386605 hasConceptScore W3011386605C54926389 @default.
- W3011386605 hasConceptScore W3011386605C62520636 @default.
- W3011386605 hasConceptScore W3011386605C76155785 @default.
- W3011386605 hasLocation W30113866051 @default.
- W3011386605 hasOpenAccess W3011386605 @default.
- W3011386605 hasPrimaryLocation W30113866051 @default.
- W3011386605 hasRelatedWork W11222364 @default.
- W3011386605 hasRelatedWork W11353818 @default.
- W3011386605 hasRelatedWork W12208184 @default.
- W3011386605 hasRelatedWork W14719770 @default.
- W3011386605 hasRelatedWork W14991473 @default.
- W3011386605 hasRelatedWork W2508318 @default.
- W3011386605 hasRelatedWork W6725010 @default.
- W3011386605 hasRelatedWork W7685913 @default.
- W3011386605 hasRelatedWork W7991636 @default.
- W3011386605 hasRelatedWork W9359371 @default.
- W3011386605 isParatext "false" @default.
- W3011386605 isRetracted "false" @default.
- W3011386605 magId "3011386605" @default.
- W3011386605 workType "article" @default.