Hierarchical pitch target model for Mandarin speech

Zhang, Zhiping; Wang, Xinhao; Yu, Yansuo; Wu, Xihong

doi:10.1109/iscslp.2010.5684865

Cited by 3 publications

(1 citation statement)

References 10 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In this way, the articulatory dynamics of F0 production becomes an integral part of both the learning and synthesis processes. Some previous research [25][26][27][28] also experimented on the same articulatory F0 production model as used here via various approaches. However, the model was used in either a hierarchical structure or a post-filtering way but seldom used on its own.…”

Section: Introductionmentioning

confidence: 99%

Model-Based Parametric Prosody Synthesis with Deep Neural Network

Líu

et al. 2016

Interspeech 2016

View full text Add to dashboard Cite

Conventional statistical parametric speech synthesis (SPSS) captures only frame-wise acoustic observations and computes probability densities at HMM state level to obtain statistical acoustic models combined with decision trees, which is therefore a purely statistical data-driven approach without explicit integration of any articulatory mechanisms found in speech production research. The present study explores an alternative paradigm, namely, model-based parametric prosody synthesis (MPPS), which integrates dynamic mechanisms of human speech production as a core component of F0 generation. In this paradigm, contextual variations in prosody are processed in two separate yet integrated stages: linguistic to motor, and motor to acoustic. Here the motor model is target approximation (TA), which generates syllable-sized F0 contours with only three motor parameters that are associated to linguistic functions. In this study, we simulate this two-stage process by linking the TA model to a deep neural network (DNN), which learns the "linguistic-motor" mapping given the "motoracoustic" mapping provided by TA-based syllable-wise F0 production. The proposed prosody modeling system outperforms the HMM-based baseline system in both objective and subjective evaluations.

show abstract

Section: Introductionmentioning

confidence: 99%

Model-Based Parametric Prosody Synthesis with Deep Neural Network

Líu

et al. 2016

Interspeech 2016

View full text Add to dashboard Cite

show abstract

Multi-speaker prosodic instance selection for HMM-based speech synthesis

Zhu

2013

2013 IEEE China Summit and International Conference on Signal and Information Processing

View full text Add to dashboard Cite

Effect of F0 contour on perception of Mandarin Chinese speech against masking

2019

PLoS ONE

View full text Add to dashboard Cite

Intonation has many perceptually significant functions in language that contribute to speech recognition. This study aims to investigate whether intonation cues affect the unmasking of Mandarin Chinese speech in the presence of interfering sounds. Specifically, intelligibility of multi-tone Mandarin Chinese sentences with maskers consisting of either two-talker speech or steady-state noise was measured in three (flattened, typical, and exaggerated) intonation conditions. Different from most of the previous studies, the present study only manipulate and modify the intonation information but preserve tone information. The results showed that recognition of the final keywords in multi-tone Mandarin Chinese sentences was much better under the original F0 contour condition than the decreased F0 contour or exaggerated F0 contour conditions whenever there was a noise or speech masker, and an exaggerated F0 contour reduced the intelligibility of Mandarin Chinese more under the speech masker condition than that under the noise masker condition. These results suggested that speech in a tone language (Mandarin Chinese) is harder to understand when the intonation is unnatural, even if the tone information is preserved, and an unnatural intonation contour decreases releasing Mandarin Chinese speech from masking, especially in a multi-person talking environment.

show abstract

Hierarchical pitch target model for Mandarin speech

Cited by 3 publications

References 10 publications

Model-Based Parametric Prosody Synthesis with Deep Neural Network

Model-Based Parametric Prosody Synthesis with Deep Neural Network

Multi-speaker prosodic instance selection for HMM-based speech synthesis

Effect of F0 contour on perception of Mandarin Chinese speech against masking

Contact Info

Product

Resources

About