Unlock instant, AI-driven research and patent intelligence for your innovation.

A Robust Speaker Recognition Method Based on Competitive Neural Network

A competitive neural network and speaker recognition technology, applied in the field of robust speaker recognition, can solve the problem of environmental noise affecting the accuracy of speaker recognition

Active Publication Date: 2021-05-28
BEIJING UNIV OF POSTS & TELECOMM
View PDF6 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] In order to solve the problem that environmental noise affects the accuracy of speaker recognition, the present invention provides a robust speaker recognition method based on competitive neural network

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 Robust Speaker Recognition Method Based on Competitive Neural Network
  • A Robust Speaker Recognition Method Based on Competitive Neural Network
  • A Robust Speaker Recognition Method Based on Competitive Neural Network

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0027] Specific embodiments of the present invention will be described in detail below in conjunction with the accompanying drawings.

[0028] figure 2 It is a flowchart of the present invention, wherein the solid line represents the direction of the training part of the process, and the dotted line represents the direction of the identification part of the process, including the following steps:

[0029] Step 1: Train a noise-invariant feature extractor. The competitive neural network is trained by extracting the features of Mel cepstrum coefficients from the training data containing noise. After the training is completed, the encoding network of the lower layer of the competition network is extracted as a feature extractor for noise invariant feature extraction.

[0030] Step 2: Train the Universal Background Model (UBM). Using a large number of background speech irrelevant to the speaker to be recognized to extract cepstral coefficient features, the feature extractor ob...

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 discloses a robust speaker recognition method based on a competitive neural network. The method uses the competitive neural network to extract acoustic features with noise invariance, and uses the features to train the speaker recognition system based on the GMM-UBM model. Competing neural network, and using the encoding network to extract noise invariant features, and then using the extracted features to implement the speaker based on the GMM‑UBM model. When training the competition network, the encoding network and the distinguishing network are trained separately. When training the encoding network, all inputs use the same clean voice label. When training the distinguishing network, the noise type of the training voice is used as the training label. Using the embodiments of the present invention, text-independent The speaker identification rate has great practical value.

Description

technical field [0001] The invention belongs to the field of voiceprint recognition and emphatically describes a robust speaker recognition method based on a competitive neural network. Background technique [0002] Speaker recognition is a technology for computers to use the information contained in speech fragments that can reflect the characteristics of the speaker to identify the identity of the speaker. This technology has very important research and application value in the fields of information security and remote identity authentication. [0003] In practical applications, the existence of environmental noise will greatly reduce the accuracy of speaker recognition. Although the commonly used speech enhancement methods can remove the noise in the speech, it will also destroy the speaker-related information in the speech while denoising. The information is not suitable for the task of speaker recognition. Therefore, directly extracting an acoustic feature that is inva...

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): G10L17/02G10L17/20G10L21/0208G10L25/24G10L25/30
CPCG10L17/02G10L17/20G10L21/0208G10L25/24G10L25/30
Inventor 于泓马占宇司中威郭军
Owner BEIJING UNIV OF POSTS & TELECOMM