Quantitative f0 contour generating device and method, and model learning device and method for f0 contour generation

Patent metadata

Field	Value
Publication number	US-2016189705-A1
Application number	US-201414911189-A
Country	US
Kind code	A1
Filing date	Aug 13, 2014
Priority date	Aug 23, 2013
Publication date	Jun 30, 2016
Grant date	—

Abstract

[Object] An object is to provide an F0 contour synthesizing device based on statistic model, to clarify correspondence between linguistic information and F0 contour while maintaining accuracy.
[Solution] An HMM learning device includes: a parameter estimating unit representing an F0 contour 133 fitting a continuous F0 contour 132 as a sum of phrase components and accent components and estimating target points of these; and an HMM learning means conducting learning of HMM 139 using the fitted F0 contour as training data. The continuous F0 contour may be decomposed to accent components 134 , phrase components 136 and micro-prosody components 138 , and separate HMMs 140, 142 and 144 may be trained. Using results of text analysis, accent components, phrase components and micro-prosody components are separately synthesized from HMMs 140, 142 and 144 and the results are synthesized to obtain an F0 contour.

First claim

1 . A quantitative F0 contour generating device, comprising:
means for generating, for an accent phrase of an utterance obtained by text analysis, accent components of an F0 contour using a given number of target points; means for generating phrase components of the F0 contour using a limited number of target points, by dividing the utterance to groups each including one or more accent phrases, in accordance with linguistic information including an utterance structure; and means for generating an F0 contour based on said accent components and said phrase components.

2 . A quantitative F0 contour generating method, comprising the steps of:
generating, for an accent phrase of an utterance obtained by text analysis, accent components of an F0 contour using a given number of target points; generating phrase components of the F0 contour using a limited number of target points, by dividing the utterance to groups each including one or more accent phrases, in accordance with linguistic information including an utterance structure; and generating an F0 contour based on said accent components and said phrase components.

3 .- 4 . (canceled)

5 . A model learning device for F0 contour generation, comprising:
F0 contour extracting means for extracting an F0 contour from a speech data signal; parameter estimating means for estimating target parameters representing phrase components and target parameters representing accent components, for representing an F0 contour fitting the extracted F0 contour by superposition of phrase components and accent components; and model learning means, performing F0 generation model learning, using a continuous F0 contour represented by the target parameters of phrase components and the target parameters of accent components estimated by said parameter estimating means as training data.

6 . The model learning device according to claim 5 , wherein
said F0 generation model includes a generation model for generating phrase components and a generation model for generating accent components; and said model learning means includes means for performing learning of said generation model for generating phrase components and said generation model for generating accent components, respectively using, as training data, a time change contour of phrase components represented by target parameters of the phrase components and a time change contour of accent components represented by target parameters of the accent components, estimated by said parameter estimating means.

7 . A model learning method for F0 contour generation, comprising the steps of:
F0 contour extracting step of extracting an F0 contour from a speech data signal; parameter estimating step of estimating target parameters representing phrase components and target parameters representing accent components, for representing an F0 contour fitting the extracted F0 contour by superposition of phrase components and accent components; and model learning step of performing F0 generation model learning, using a continuous F0 contour represented by the target parameters of phrase components and the target parameters of accent components estimated by said parameter estimating means as training data.

8 . The model learning method according to claim 7 , wherein
said F0 generation model includes a generation model for generating phrase components and a generation model for generating accent components; and
said model learning step includes the step of performing learning of said generation model for generating phrase components and said generation model for generating accent components, respectively using, as training data, a time change contour of phrase components represented by target parameters of the phrase components and a time change contour of accent components represented by target parameters of the accent components, estimated at said parameter estimating step.

Assignees

• Nat Inst Inf & Comm Tech
• Nat Inst Inf & Comm Tech

Inventors

• Ni Jinfu
• Shiga Yoshinori

Classifications

• G10L25/18

  the extracted parameters being spectral information of each sub-band · CPC title

• G10L13/10 Primary

  Prosody rules derived from text; Stress or intonation · CPC title

• G10L13/027

  Concept to speech synthesisers; Generation of natural phrases from machine-based concepts (generation of parameters for speech synthesis out of text G10L13/08) · CPC title

• G10L13/086

  Detection of language · CPC title

• G10L21/0205

  Physics · mapped topic