Text independence based short speech speaker confirmation method

A speaker confirmation and text-independent technology, applied in speech analysis, instruments, etc., can solve the problems of reducing the amount of calculation, reducing the performance of system recognition, and reducing the amount of calculation of the recognition system, so as to shorten the recognition time, reduce the error rate, and system The best performance

Inactive Publication Date: 2019-05-07
CHINA CHANGFENG SCI TECH IND GROUPCORP
View PDF0 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

One is the core selection algorithm of Gaussian components. This algorithm organizes the Gaussian components of each UBM into a tree structure, and quickly selects several Gaussian components with high similarity with the test voice frame through the tree structure in the speech test stage. This method reduces At the same time, because the Gaussian component search may not be able to find the first few Gaussian components with the highest true likelihood score, the system identification performance will be reduced to a certain extent.
The other method is to quickly calculate the likelihood score, that is, for a frame of speech, first calculate the score of each Gaussian component of the frame of speech in UBM, and select the first C components for marking according to the method of high score priority. When the test voice is calculated on the personal GMM, only the likelihood scores under the C Gaussian components corresponding to the UBM in the GMM model are calculated. Therefore, this method can reduce the calculation amount of the system during the voice test and make the recognition result as soon as possible. However, , the model still needs to build a high-level voiceprint model GMM for each speaker based on UBM during voice training, and still needs a lot of calculations in the training model stage

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
  • Text independence based short speech speaker confirmation method
  • Text independence based short speech speaker confirmation method
  • Text independence based short speech speaker confirmation method

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0016] The present invention mainly aims at the problems of the large amount of calculation in the UBM-MAP-GMM model and the influence of some Gaussian components on the recognition performance, and proposes a recognition method based on UBM-CM-MAP-GMM. This method is mainly based on UBM-MAP- In the GMM system, there are situations in which the speaker's speech training is insufficient and thus affects the final judgment result. The Gaussian component of the UBM model is screened, and a targeted low-order UBM is established for each speaker, that is, the competitor model CM, and then based on The low-level UBM uses the training voice to build its own voiceprint model GMM for each speaker. The main purpose of this recognition method is to effectively improve the performance of segment speech speaker recognition, and reduce the time consumption of speaker recognition in the speech test stage on the basis of reducing the mixing degree of CM and GMM.

[0017] The main idea of ​​th...

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 relates to a text independence based short speech speaker confirmation method. Based on a UBM-CM-MAP-GMM model, as low-score Gauss components have negative effects on the recognition performance of a voiceprint recognition system, low-score Gauss components are deleted, high-score Gauss components are selected, the high-score Gauss components are combined, a competitor model CM is thus built for each target speaker, and based on the model CM of each speaker, a voiceprint model GMM of the speaker is obtained through MAP self-adaption.

Description

technical field [0001] The invention relates to voiceprint recognition technology in the technical fields of artificial intelligence and pattern recognition, in particular to a text-independent short voice speaker confirmation method. Background technique [0002] Voiceprint recognition technology is a comprehensive subject that combines computer, biostatistics, biosensors and other disciplines. This technology mainly identifies people's identities through each person's unique innate physiological characteristics or acquired behavioral characteristics. . Voiceprint Recognition (Voiceprint Recognition), also known as speaker recognition, is a biometric technology that automatically identifies the identity of the speaker through the characteristic information contained in the human voice. [0003] In the actual application process of voiceprint recognition technology, we often face the situation that the voice data is short and the amount of data is scarce. The impact of the...

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): G10L17/00G10L17/04
Inventor 杨瑞瑞柴秀英
Owner CHINA CHANGFENG SCI TECH IND GROUPCORP
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