Speaker embedding learning method and speaker recognition method and system

A learning method and learning system technology, applied in the speaker identification method and system, the speaker embedding learning method field, can solve the problem of no good and so on

Active Publication Date: 2020-10-23
AISPEECH CO LTD
View PDF2 Cites 2 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, despite a powerful DNN front end, d-vectors stil

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
  • Speaker embedding learning method and speaker recognition method and system
  • Speaker embedding learning method and speaker recognition method and system
  • Speaker embedding learning method and speaker recognition method and system

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0026] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments It is a part of embodiments of the present invention, but not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.

[0027] It should be noted that, in the case of no conflict, the embodiments in the present application and the features in the embodiments can be combined with each other.

[0028] The invention may be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, progr...

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 speaker embedding learning method, which comprises the steps of performing frame-level feature extraction on a speaker voice segment to obtain a plurality of frame-level depth features, performing second-order pooling processing on the plurality of frame-level depth features to obtain segment-level depth features, and converting the segment-level depth features into segment-level speaker embedding by adopting an affine layer. According to the embodiment of the invention, a second-order pooling processing mode is adopted when the frame-level depth features are converted into the segment-level depth features, therefore, not only is the frame-level speaker feature considered, but also the dynamic information in the voice segment is considered, so that the finally learned speaker embedding can reflect the speaker feature more accurately, the quality of the speaker embedding feature is improved, and the accuracy and reliability of a speaker recognition task performed on the basis of the speaker embedding feature are facilitated.

Description

technical field [0001] The invention relates to the technical field of artificial intelligence, in particular to a speaker embedding learning method, a speaker identification method and a system. Background technique [0002] Currently, deep speaker embeddings are the main approach for speaker identity modeling. Unlike shallow models (e.g., Gaussian mixture models (GMM) or factor analysis), deep neural networks (DNNs) show incredible nonlinear modeling capabilities for complex data distributions. Therefore, one of the hottest topics is representation learning using DNNs, which aims to learn highly compact and informative embeddings to represent raw inputs. [0003] In the field of speaker recognition, the d-vector paradigm is the first well-known DNN based on an embedding learning framework, which uses a speaker discrimination DNN to extract frame-level deep features, which are then averaged into individual speaker embeddings. However, despite the powerful DNN front-end, d...

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): G10L15/02G10L15/16
CPCG10L15/02G10L15/16
Inventor 俞凯王帅杨叶新钱彦旻
Owner AISPEECH 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