10th ISCA Workshop on Speech Synthesis (SSW 10)最新文献

Implementation of DNN-based real-time voice conversion and its improvements by audio data augmentation and mask-shaped device 基于dnn的实时语音转换的实现及其音频数据增强和掩模器件的改进

10th ISCA Workshop on Speech Synthesis (SSW 10) Pub Date : 2019-09-20 DOI: 10.21437/ssw.2019-17

Riku Arakawa, Shinnosuke Takamichi, H. Saruwatari

引用次数: 19

How to train your fillers: uh and um in spontaneous speech synthesis 如何在自发语音合成中训练你的填充语:呃和嗯

10th ISCA Workshop on Speech Synthesis (SSW 10) Pub Date : 2019-09-20 DOI: 10.21437/ssw.2019-44

Éva Székely, G. Henter, J. Beskow, Joakim Gustafson

{"title":"How to train your fillers: uh and um in spontaneous speech synthesis","authors":"Éva Székely, G. Henter, J. Beskow, Joakim Gustafson","doi":"10.21437/ssw.2019-44","DOIUrl":"https://doi.org/10.21437/ssw.2019-44","url":null,"abstract":"Using spontaneous conversational speech for TTS raises questions on how disﬂuencies such as ﬁlled pauses (FPs) should be approached. Detailed annotation of FPs in training data enables precise control at synthesis time; coarse or nonexistent FP annotation, when combined with stochastic attention-based neural TTS, leads to synthesisers that insert these phenomena into ﬂuent prompts on their own accord. In this study we investigate, objectively and subjectively, the effects of FP annotation and the impact of relinquishing control over FPs in a Tacotron TTS system. The training corpus comprised 9 hours of single-speaker breath groups extracted from a conversational podcast. Systems trained with no or location-only FP annotation were found to reproduce FP locations and types (uh/um) in a pat-tern broadly similar to that of the corpus. We also studied the effect of FPs on natural and synthetic speech rate and the interchangeability of FP types. Interestingly, subjective tests indicate that synthesiser-predicted FP types from location-only annotation often were preferred over specifying the ground-truth type. In contrast, a more precise annotation, allowing us to focus training on the most ﬂuent parts of the corpus, improved rated naturalness when synthesising ﬂuent speech.","PeriodicalId":418150,"journal":{"name":"10th ISCA Workshop on Speech Synthesis (SSW 10)","volume":"97 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-09-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"124665293","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 25

Prosody Prediction from Syntactic, Lexical, and Word Embedding Features 从句法、词汇和词嵌入特征预测韵律

10th ISCA Workshop on Speech Synthesis (SSW 10) Pub Date : 2019-09-20 DOI: 10.21437/ssw.2019-48

Rose Sloan, S. S. Akhtar, Bryan Li, Ritvik Shrivastava, A. Gravano, Julia Hirschberg

引用次数: 8

Unsupervised Learning of a Disentangled Speech Representation for Voice Conversion 一种用于语音转换的解纠缠语音表示的无监督学习

10th ISCA Workshop on Speech Synthesis (SSW 10) Pub Date : 2019-09-20 DOI: 10.21437/ssw.2019-15

Tobias Gburrek, Thomas Glarner, Janek Ebbers, R. Haeb-Umbach, P. Wagner

{"title":"Unsupervised Learning of a Disentangled Speech Representation for Voice Conversion","authors":"Tobias Gburrek, Thomas Glarner, Janek Ebbers, R. Haeb-Umbach, P. Wagner","doi":"10.21437/ssw.2019-15","DOIUrl":"https://doi.org/10.21437/ssw.2019-15","url":null,"abstract":"This paper presents an approach to voice conversion, whichdoes neither require parallel data nor speaker or phone labels fortraining. It can convert between speakers which are not in thetraining set by employing the previously proposed concept of afactorized hierarchical variational autoencoder. Here, linguisticand speaker induced variations are separated upon the notionthat content induced variations change at a much shorter timescale, i.e., at the segment level, than speaker induced variations,which vary at the longer utterance level. In this contribution wepropose to employ convolutional instead of recurrent networklayers in the encoder and decoder blocks, which is shown toachieve better phone recognition accuracy on the latent segmentvariables at frame-level due to their better temporal resolution.For voice conversion the mean of the utterance variables is re-placed with the respective estimated mean of the target speaker.The resulting log-mel spectra of the decoder output are used aslocal conditions of a WaveNet which is utilized for synthesis ofthe speech waveforms. Experiments show both good disentan-glement properties of the latent space variables, and good voiceconversion performance.","PeriodicalId":418150,"journal":{"name":"10th ISCA Workshop on Speech Synthesis (SSW 10)","volume":"27 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-09-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"125235313","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 3

Low computational cost speech synthesis based on deep neural networks using hidden semi-Markov model structures 基于隐藏半马尔可夫模型结构的深度神经网络低计算成本语音合成

10th ISCA Workshop on Speech Synthesis (SSW 10) Pub Date : 2019-09-20 DOI: 10.21437/ssw.2019-32

Motoki Shimada, Kei Hashimoto, Keiichiro Oura, Yoshihiko Nankaku, K. Tokuda

{"title":"Low computational cost speech synthesis based on deep neural networks using hidden semi-Markov model structures","authors":"Motoki Shimada, Kei Hashimoto, Keiichiro Oura, Yoshihiko Nankaku, K. Tokuda","doi":"10.21437/ssw.2019-32","DOIUrl":"https://doi.org/10.21437/ssw.2019-32","url":null,"abstract":"We propose a method of changing the units of input features from states used conventionally to phonemes and moras to reduce the computational cost of deep neural networks (DNNs) with a hidden semi-Markov model structure for speech synthesis, which can model acoustic features and a temporal structure in a uniﬁed framework. Neural networks with very deep and wide structures have recently been applied successfully in the ﬁeld of speech synthesis. However, such models have very high computational cost, so they are not being applied on platforms with limited resources. To solve this problem, we increased the length of time of DNN input units. We used phoneme or mora units, which are longer than the state units used conventionally. Increasing the length in time of units of input features reduces the number of DNN forward propagations required for speech synthesis, reducing the computational cost. Since a mora in Japanese exhibits isochronism, the duration can be represented more appropriately than the phoneme units expressing consonants and vowels of different lengths with one neural network. Experimental results indicate that compared with speech synthesis based on a DNN with frame inputs, computational cost can be reduced by 97% without degrading the naturalness of the synthesized speech with the proposed method.","PeriodicalId":418150,"journal":{"name":"10th ISCA Workshop on Speech Synthesis (SSW 10)","volume":"36 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-09-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"132002206","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 0

GlottDNN-based spectral tilt analysis of tense voice emotional styles for the expressive 3D numerical synthesis of vowel [a] 基于glottdnn的紧张语音情感类型的频谱倾斜分析及其在元音三维数值合成中的应用[a]

10th ISCA Workshop on Speech Synthesis (SSW 10) Pub Date : 2019-09-20 DOI: 10.21437/ssw.2019-24

M. Freixes, M. Arnela, Francesc Alías, J. Socoró

{"title":"GlottDNN-based spectral tilt analysis of tense voice emotional styles for the expressive 3D numerical synthesis of vowel [a]","authors":"M. Freixes, M. Arnela, Francesc Alías, J. Socoró","doi":"10.21437/ssw.2019-24","DOIUrl":"https://doi.org/10.21437/ssw.2019-24","url":null,"abstract":"Three-dimensional (3D) acoustic models allow for an accu-rate modelling of acoustic wave propagation in 3D realistic vocal tracts. However, voice generated by these approaches is still limited in terms of expressiveness, which could be improved through proper modiﬁcations of the glottal source excitation. This work aims at adding some expressiveness to a 3D numerical synthesis approach based on the Finite Element Method (FEM) that uses as input an LF (Liljencrants-Fant) model controlled by the glottal shape parameter R d . To that effect, a parallel Spanish speech corpus containing neutral and tense voice emotional styles is analysed with the GlottDNN vocoder, ob-taining F 0 and spectral tilt parameters associated with the glottal excitation. The variations of these two parameters are computed for happy and aggressive styles with reference to neutral speech, differentiating between stressed and unstressed vowels [a]. From this analysis, F 0 and R d values are then derived and used in the LF-FEM based synthesis of vowels [a] to resemble the aforementioned expressive styles. Results show that it is necessary to increase F 0 and decrease R d with respect to neutral speech, with larger deviations for happy than aggressive style, especially for the stressed vowels.","PeriodicalId":418150,"journal":{"name":"10th ISCA Workshop on Speech Synthesis (SSW 10)","volume":"17 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-09-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"117011581","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 3

Initial investigation of encoder-decoder end-to-end TTS using marginalization of monotonic hard alignments 基于单调硬对准边缘化的编码器-解码器端到端TTS初步研究

10th ISCA Workshop on Speech Synthesis (SSW 10) Pub Date : 2019-09-20 DOI: 10.21437/ssw.2019-38

Yusuke Yasuda, Xin Wang, J. Yamagishi

引用次数: 7

Building Multilingual End-to-End Speech Synthesisers for Indian Languages 为印度语言建立多语言端到端语音合成器

10th ISCA Workshop on Speech Synthesis (SSW 10) Pub Date : 2019-09-20 DOI: 10.21437/ssw.2019-35

A. Prakash, Anju Leela Thomas, S. Umesh, H. Murthy

{"title":"Building Multilingual End-to-End Speech Synthesisers for Indian Languages","authors":"A. Prakash, Anju Leela Thomas, S. Umesh, H. Murthy","doi":"10.21437/ssw.2019-35","DOIUrl":"https://doi.org/10.21437/ssw.2019-35","url":null,"abstract":"Building text-to-speech (TTS) synthesisers is a difﬁcult task, especially for low resource languages. Language-speciﬁc modules need to be developed for system building. End-to-end speech synthesis has become a popular paradigm as a TTS can be trained using only < text, audio > pairs. However, end-to-end speech synthesis is not scalable in a multilanguage scenario, as the vocabulary increases with the number of different scripts. In this paper, TTSes are trained for Indian languages using two text representations– character-based and phone-based. For the character-based approach, a multi-language character map (MLCM) is proposed to easily train Indic speech synthesisers. The phone-based approach uses the common label set (CLS) representation for Indian languages. Both approaches leverage the similarities that exist among the languages. The advantage is a compact representation across multiple languages. Experiments are conducted by building TTSes using monolingual data and by pooling data across two languages. The ability to syn-thesise code-mixed text using the phone-based approach is also assessed. Subjective evaluations indicate that reasonably good quality Indic TTSes can be developed using both approaches. This emphasises the need to incorporate multilingual text processing in the end-to-end framework.","PeriodicalId":418150,"journal":{"name":"10th ISCA Workshop on Speech Synthesis (SSW 10)","volume":"22 1","pages":"0"},"PeriodicalIF":0.0,"publicationDate":"2019-09-20","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"123629396","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}

引用次数: 24

A Comparison of Recent Neural Vocoders for Speech Signal Reconstruction 几种用于语音信号重构的神经声码器的比较

10th ISCA Workshop on Speech Synthesis (SSW 10) Pub Date : 2019-09-20 DOI: 10.21437/ssw.2019-2

Prachi Govalkar, Johannes Fischer, Frank Zalkow, C. Dittmar

引用次数: 27

Impacts of input linguistic feature representation on Japanese end-to-end speech synthesis 输入语言特征表示对日语端到端语音合成的影响

10th ISCA Workshop on Speech Synthesis (SSW 10) Pub Date : 2019-09-20 DOI: 10.21437/ssw.2019-30

Takato Fujimoto, Kei Hashimoto, Keiichiro Oura, Yoshihiko Nankaku, K. Tokuda

引用次数: 17