Matches in SemOpenAlex for { <https://semopenalex.org/work/W2793399581> ?p ?o ?g. }
Showing items 1 to 84 of
84
with 100 items per page.
- W2793399581 endingPage "172" @default.
- W2793399581 startingPage "161" @default.
- W2793399581 abstract "Abstract This paper presents a method to convert articulatory movements into speech waveforms using a data-driven approach. In this method, recorded electromagnetic midsagittal articulography (EMA) measurements are converted into both spectral features (i.e., Mel-cepstra) and excitation features (i.e., power, voiced/unvoiced flag, and F0) from which speech waveforms are then reconstructed. By considering the nonlinear and dynamic dependency relationships between articulatory movements and acoustic signals, this study adopts bidirectional long short-term memory (BLSTM) based recurrent neural networks (RNN) for the articulatory-to-acoustic conversion. Due to the limitations of current data acquisition technology, recorded articulatory movements are inevitably insufficient to completely describe the articulatory configuration during pronunciation. Therefore, this paper proposes to further augment the model input by concatenating EMA vectors with two other representations. First, the posterior probabilities derived from a phoneme classifier are concatenated with EMA features to provide a linguistic description of each frame for acoustic feature prediction. The classifier is trained to determine the phoneme label of each frame based on the observed EMA features. Second, a cascaded prediction strategy is designed to utilize the predicted spectral features as auxiliary input to boost the prediction accuracy of the excitation features. The results of experiment show that BLSTM-RNNs can achieve a better objective and subjective performance than deep neural networks (DNN) and Gaussian mixture models (GMM) in articulatory-to-acoustic conversion. In addition, our results show that the proposed methods for integrating linguistic representation and utilizing a cascaded prediction strategy can further improve the accuracy of acoustic feature prediction." @default.
- W2793399581 created "2018-03-29" @default.
- W2793399581 creator A5002028345 @default.
- W2793399581 creator A5057227915 @default.
- W2793399581 creator A5059767940 @default.
- W2793399581 date "2018-05-01" @default.
- W2793399581 modified "2023-10-16" @default.
- W2793399581 title "Articulatory-to-acoustic conversion using BLSTM-RNNs with augmented input representation" @default.
- W2793399581 cites W1977591085 @default.
- W2793399581 cites W1982854652 @default.
- W2793399581 cites W2008120082 @default.
- W2793399581 cites W2020304287 @default.
- W2793399581 cites W2023407680 @default.
- W2793399581 cites W2059169992 @default.
- W2793399581 cites W2064675550 @default.
- W2793399581 cites W2068056889 @default.
- W2793399581 cites W2078528584 @default.
- W2793399581 cites W2100495367 @default.
- W2793399581 cites W2120605154 @default.
- W2793399581 cites W2135707066 @default.
- W2793399581 cites W2150355110 @default.
- W2793399581 cites W2156142001 @default.
- W2793399581 cites W2160815625 @default.
- W2793399581 cites W2260035229 @default.
- W2793399581 cites W2508425411 @default.
- W2793399581 cites W2518315489 @default.
- W2793399581 cites W4206144308 @default.
- W2793399581 doi "https://doi.org/10.1016/j.specom.2018.02.008" @default.
- W2793399581 hasPublicationYear "2018" @default.
- W2793399581 type Work @default.
- W2793399581 sameAs 2793399581 @default.
- W2793399581 citedByCount "9" @default.
- W2793399581 countsByYear W27933995812018 @default.
- W2793399581 countsByYear W27933995812019 @default.
- W2793399581 countsByYear W27933995812020 @default.
- W2793399581 countsByYear W27933995812021 @default.
- W2793399581 countsByYear W27933995812022 @default.
- W2793399581 countsByYear W27933995812023 @default.
- W2793399581 crossrefType "journal-article" @default.
- W2793399581 hasAuthorship W2793399581A5002028345 @default.
- W2793399581 hasAuthorship W2793399581A5057227915 @default.
- W2793399581 hasAuthorship W2793399581A5059767940 @default.
- W2793399581 hasConcept C147168706 @default.
- W2793399581 hasConcept C153180895 @default.
- W2793399581 hasConcept C154945302 @default.
- W2793399581 hasConcept C197424946 @default.
- W2793399581 hasConcept C23224414 @default.
- W2793399581 hasConcept C28490314 @default.
- W2793399581 hasConcept C41008148 @default.
- W2793399581 hasConcept C50644808 @default.
- W2793399581 hasConcept C554190296 @default.
- W2793399581 hasConcept C76155785 @default.
- W2793399581 hasConcept C95623464 @default.
- W2793399581 hasConceptScore W2793399581C147168706 @default.
- W2793399581 hasConceptScore W2793399581C153180895 @default.
- W2793399581 hasConceptScore W2793399581C154945302 @default.
- W2793399581 hasConceptScore W2793399581C197424946 @default.
- W2793399581 hasConceptScore W2793399581C23224414 @default.
- W2793399581 hasConceptScore W2793399581C28490314 @default.
- W2793399581 hasConceptScore W2793399581C41008148 @default.
- W2793399581 hasConceptScore W2793399581C50644808 @default.
- W2793399581 hasConceptScore W2793399581C554190296 @default.
- W2793399581 hasConceptScore W2793399581C76155785 @default.
- W2793399581 hasConceptScore W2793399581C95623464 @default.
- W2793399581 hasLocation W27933995811 @default.
- W2793399581 hasOpenAccess W2793399581 @default.
- W2793399581 hasPrimaryLocation W27933995811 @default.
- W2793399581 hasRelatedWork W1519502414 @default.
- W2793399581 hasRelatedWork W1549289070 @default.
- W2793399581 hasRelatedWork W2148757832 @default.
- W2793399581 hasRelatedWork W2151447942 @default.
- W2793399581 hasRelatedWork W2293457016 @default.
- W2793399581 hasRelatedWork W2611614995 @default.
- W2793399581 hasRelatedWork W2800726910 @default.
- W2793399581 hasRelatedWork W2900389025 @default.
- W2793399581 hasRelatedWork W2902723393 @default.
- W2793399581 hasRelatedWork W3107474891 @default.
- W2793399581 hasVolume "99" @default.
- W2793399581 isParatext "false" @default.
- W2793399581 isRetracted "false" @default.
- W2793399581 magId "2793399581" @default.
- W2793399581 workType "article" @default.