Voice conversion method facing to multi-time scale prosodic features

A multi-time scale and prosodic feature technology, applied in speech analysis, speech recognition, speech synthesis, etc., can solve problems such as correlation

Inactive Publication Date: 2013-04-03
NANJING UNIV OF POSTS & TELECOMM
View PDF1 Cites 16 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] In the study of prosodic characteristics, the traditional view is that prosodic characteristics refer to the supersegmental characteristics, that is, the time evolution characteristics of speech signals at the supersegmental level, and that this characteristic is related to the arrangement order of speech segments on the time axis However, there is a correlation between prosodic characteristics and segmental characteristics in essence, so this view has certain limitations

Method used

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
View more

Image

Smart Image Click on the blue labels to locate them in the text.
Viewing Examples
Smart Image
  • Voice conversion method facing to multi-time scale prosodic features
  • Voice conversion method facing to multi-time scale prosodic features
  • Voice conversion method facing to multi-time scale prosodic features

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0051] specific implementation plan

[0052] Below in conjunction with accompanying drawing, the implementation of technical scheme is described in further detail:

[0053] Such as figure 1 , the present invention is based on the double hidden Markov model multi-time scale prosody feature conversion method, the steps are as follows:

[0054] The first step is to pre-process the speech signals of the input source speaker and target speaker, such as pre-emphasis, framing, and windowing, such as figure 2 As shown, according to the grammatical rules of the speech signal and the auditory perception characteristics of the human ear, a sentence can be decomposed into several phrases, and these phrases can completely and independently express a semantic meaning. A phrase can be divided into several syllables, and each syllable is the basic unit of pronunciation. The different prosodic characteristics of speech signals are best represented at different time scales. Speech is divide...

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

PUM

No PUM Login to view more

Abstract

The invention discloses a voice conversion method facing to multi-time scale prosodic features and belongs to the technical field of voice signal processing. The method comprises firstly conducting prosodic feature analysis and parameterization extraction on voice signals under the multi-time scale; then building a alternative model for the multi-time scale prosodic features based on a double-hidden markov model; and finally forming estimation features of a target speaker in a conversion stage to obtain converted voice. The voice conversion method can meticulous and complete depict the prosodic features from whole to local, overcomes fuzziness and complexity of prosodic information expression, achieves prosodic feature conversion by building a time sequence statistic model, strengthens individual information of the speaker of converted voice, and simultaneously improves speech intelligibility naturalness of the converted voice.

Description

technical field [0001] The invention relates to a speech conversion technology, in particular to a speech conversion method based on the multi-time-scale prosody feature of a double hidden Markov model, and belongs to the technical field of speech signal processing. Background technique [0002] Speech conversion is an emerging research branch in the field of speech signal processing in recent years. It is based on the research of speaker recognition and speech synthesis, and it is also the enrichment and extension of the connotation of these two branches. [0003] The goal of speech conversion is to change the personality feature information in the source speaker's voice, so that it has the personality characteristics of the target speaker, so that the converted voice sounds like the target speaker's voice, while the semantic information in it remains unchanged. Change. [0004] A speech conversion system with good performance must not only maintain the auditory quality of...

Claims

the structure of the environmentally friendly knitted fabric provided by the present invention; figure 2 Flow chart of the yarn wrapping machine for environmentally friendly knitted fabrics and storage devices; image 3 Is the parameter map of the yarn covering machine
Login to view more

Application Information

Patent Timeline
no application Login to view more
Patent Type & Authority Applications(China)
IPC IPC(8): G10L21/00G10L13/00G10L15/02G10L15/14
Inventor 李燕萍张玲华
Owner NANJING UNIV OF POSTS & TELECOMM
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap