Voice synthesis method and device and electronic equipment

A technology of speech synthesis and speech library, applied in speech synthesis, speech analysis, instruments, etc., can solve the problems of strong sound quality, poor stability outside the set, loss of sound quality and tone details, etc., to improve the effect and increase efficiency Effect

Active Publication Date: 2019-07-23
BEIJING SINOVOICE TECH CO LTD
View PDF15 Cites 1 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

The two have their own strengths and weaknesses, and they cannot completely replace each other: the sound selection and splicing synthesis have realistic sound quality and real sound length, but the splicing flaws are obvious and the stability outside the set is poor; the statistical parameter synthesis is stable and the synergistic pronunciation is s

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 synthesis method and device and electronic equipment
  • Voice synthesis method and device and electronic equipment
  • Voice synthesis method and device and electronic equipment

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0067] figure 1 It is a flowchart of steps of a speech synthesis method provided by an embodiment of the present invention.

[0068] refer to figure 1 As shown, the speech synthesis method provided by this embodiment is applied to electronic devices such as electronic computers or speech synthesis equipment, and specifically includes the following steps:

[0069] S1. Perform text analysis on the input text.

[0070] When the user directly inputs or other electronic equipment inputs the corresponding text, text analysis is performed on the input text, and the target primitive sequence and corresponding context information are obtained from it. The target primitive sequence here includes multiple target primitives.

[0071] S2. Utilize the traditional model decision tree to determine the subcategory number and the corresponding Gaussian distribution model respectively described in the voice selection target model of the contextual information in the speech bank.

[0072] The...

Embodiment 2

[0150] figure 2 It is a structural block diagram of a speech synthesis device provided by an embodiment of the present invention.

[0151] refer to figure 2 As shown, the speech synthesis device provided by this embodiment is applied to electronic equipment such as electronic computers or speech synthesis equipment, and specifically includes a text analysis module 10, a first calculation module 20, a distance calculation module 30, a grid construction module 40, a second Calculation module 50 , third calculation module 60 , fourth calculation module 70 , path selection module 80 and splicing output module 90 .

[0152] The text analysis module is used to perform text analysis on the input text.

[0153] When the user directly inputs or other electronic equipment inputs the corresponding text, text analysis is performed on the input text, and the target primitive sequence and corresponding context information are obtained from it. The target primitive sequence here include...

Embodiment 3

[0186] This embodiment provides an electronic device, such as a speech synthesis device, an electronic computer or a mobile terminal, etc., which is provided with the speech synthesis device provided in the previous embodiment. The device is used to perform text analysis on the input text to obtain the target primitive sequence and the corresponding context information; for the context information, the traditional model decision tree is used to determine the context information in the voice selection target model of the speech library. The subclass number and the corresponding Gaussian distribution model are used to obtain the corresponding pre-selection results; the pre-selection results are used to form a column for each target primitive in turn, and finally the target primitive sequence forms a set of candidate grids; the context information is input into the deep learning model to obtain the acoustic parameter envelope, primitive duration, and boundary frame acoustic parame...

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 embodiment of the invention provides a voice synthesis method and device and electronic equipment. According to the technical scheme, on a voice selection splicing synthesis path, a depth learningtechnology is moderately introduced, but a traditional statistical learning technology is not thoroughly abandoned, and thus the advantages of both parties are made use of; the key innovation is thata depth learning model is adopted for generating simulation data to conduct back feeding on training of a traditional statistical learning model, the effect of traditional learning is improved from two aspects of algorithms and data, and thus the effect of voice synthesis is improved.

Description

technical field [0001] The present invention relates to the technical field of speech synthesis, in particular to a speech synthesis method, device and electronic equipment. Background technique [0002] In recent years, as the wave of deep learning has swept across the related fields of machine learning, the field of speech synthesis has also been surging. From acoustic parameter modeling, speech enhancement, vocoder, to prosody analysis and other text preprocessing links, they have tried to apply State-of-the-art deep learning techniques, or even attempts to model "end-to-end" directly from text to waveforms, have achieved impressive results. [0003] In the past ten years of development in the field of speech synthesis, the contention between the two routes of statistical parameter synthesis and voice selection splicing synthesis has been maintained. The two have their own strengths and weaknesses, and they cannot completely replace each other: the sound selection and sp...

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
IPC IPC(8): G10L13/02G10L13/08G10L25/30
CPCG10L13/02G10L13/08G10L25/30
Inventor 王愈李健张连毅武卫东
Owner BEIJING SINOVOICE TECH CO LTD
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