A Speaker Labeling Method

A labeling method and speaker technology, applied in speech analysis, instruments, etc., can solve the problems of sensitive initialization and poor labeling results.

Active Publication Date: 2020-07-07
TSINGHUA UNIV
View PDF11 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

But this system is very sensitive to initialization, especially when a certain speaker is dominant in the speech to be tested, and the remaining speakers speak little speech, the VB-HMM method can easily distribute the speech to be tested evenly to each speaker, leading to poor labeling results

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
  • A Speaker Labeling Method
  • A Speaker Labeling Method
  • A Speaker Labeling Method

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0085] The speaker marking method proposed by the present invention will be further described in detail below with reference to the drawings and specific embodiments.

[0086] The speaker marking method proposed by the present invention is divided into three stages: Ivec-PLDA-AHC speaker marking stage, a priori probability initialization stage and VB-HMM speaker marking stage. The number of speakers S=2 of the voice data to be tested (to be marked) used in this embodiment. The overall process of this method is as follows figure 1 As shown, the method includes the following steps:

[0087] 1) Ivec-PLDA-AHC speaker marking stage; the specific steps are as follows:

[0088] 1-1) Obtain a piece of voice data X to be tested, and the total number of speakers S in the voice data to be tested is known (S≥2). The voice data to be tested needs to have multiple conversations, which can be telephone voice, conference voice, broadcast news voice, the data source can be recorded by yourself, or...

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 proposes a speaker marking method, and belongs to the technical field of voiceprint recognition, mode recognition and machine learning. The method includes three stages: in the first stage, voice data to be tested are divided into fragments with equal length through an i-vector probabilistic linear discriminant analysis-agglomerative hierarchical clustering method, and then the fragments are clustered into categories of which the total number is equal to that of speakers; in the second stage, prior probability that a fragment belongs to a speaker is obtained; and in the third stage, iteration is performed through a variational Bayes hidden Markov method based on soft decision, when a system converges, a speaker to which a fragment belongs is calculated, and speaker marking is finished. The speaker marking method provided by the invention combines advantages of two speaker marking methods, and can effectively improve the accuracy rate of speaker marking.

Description

Technical field [0001] The invention belongs to the technical field of voiceprint recognition, pattern recognition and machine learning, and particularly relates to a speaker marking method based on Ivec-PLDA-AHC and VB-HMM. Background technique [0002] Speaker marking is to process a segment of speech from multiple people, and to gather together the speech fragments of the same speaker, that is, to mark who is speaking when. The significance of speaker tagging is that when it is applied to conference calls and international conferences, it can be saved as meeting records. At the same time, accurate speaker identification will naturally help subsequent speech processing and semantic recognition. In addition, in the monitoring field, the speaker tag can record the voice and language of the monitored object, and it can be applied to the public security field or the military field, and contributes to the protection of public order and even national security. [0003] The existing 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
Patent Type & Authority Patents(China)
IPC IPC(8): G10L25/51G10L25/27
CPCG10L25/27G10L25/51
Inventor 陈仙红何亮徐灿刘加
Owner TSINGHUA UNIV
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