Looking for breakthrough ideas for innovation challenges? Try Patsnap Eureka!

Speaker verification system and replay attack detection method thereof

A technology of speaker verification and replay attack, applied in the field of voiceprint recognition, it can solve the problem that the discrimination effect is not optimal, and achieve the effect of improving the accuracy.

Pending Publication Date: 2020-09-01
珠海造极声音科技有限公司
View PDF4 Cites 4 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

On the other hand, most researchers widely use the balanced cross-entropy loss function to identify replay attacks, but because of the differences in the distribution of training sets, test sets, evaluation sets, and real data, the identification effect may not be optimal

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 verification system and replay attack detection method thereof
  • Speaker verification system and replay attack detection method thereof
  • Speaker verification system and replay attack detection method thereof

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0086] See figure 1 , Embodiment 1 of the present invention provides a replay attack detection method for a speaker verification system, including S1-S5;

[0087] S1: Extract the MGD-gram feature (Modified group delay, modified group delay), STFT-gram feature (STFT, short-time Fourier transform, short-time Fourier transform) and CQT-gram feature ( Constant Q transform, constant Q transform);

[0088] S2. Input the MGD-gram feature of the speech signal to be tested into a trained first end-to-end residual network model with a balanced focus loss function;

[0089] S3. Input the STFT-gram feature of the speech signal to be tested into a trained second end-to-end residual network model with a balanced focus loss function;

[0090] S4. Input the CQT-gram feature of the speech signal to be tested into a trained third end-to-end residual network model with a balanced focus loss function;

[0091] S5. Fuse the first score, the second score, and the third score in a determined fusi...

Embodiment 2

[0126] This embodiment also provides a speaker verification system, including:

[0127] In the second aspect, an embodiment of the present invention provides a replay attack detection system of a speaker verification system, including:

[0128] An extraction module is used to extract the MGD-gram feature, STFT-gram feature and CQT-gram feature of the speech signal to be tested respectively;

[0129] The first input module is used to input the MGD-gram feature of the speech signal to be tested to the trained first end-to-end residual network model with a balanced focus loss function to obtain the first score;

[0130] The second input module is used to input the STFT-gram feature of the speech signal to be tested into a trained second end-to-end residual network model with a balanced focus loss function to obtain a second score

[0131] The 3rd input module is used for inputting the CQT-gram feature of described speech signal to be tested to the 3rd end-to-end residual network...

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 speaker verification system and a replay attack detection method thereof. The method comprises the following steps: extracting an MGD-gram feature, an STFT-gram feature and a CQT-gram feature of a to-be-detected voice signal; respectively inputting the MGD-gram feature, the STFT-gram feature and the CQT-gram feature of the to-be-detected voice signal intoa first end-to-end residual network model, a second end-to-end residual network model and a third end-to-end residual network model which are trained and have a balance focus loss function; obtaininga first score, a second score and a third score; and fusing the first score, the second score and the third score in a determined fusion mode, and outputting a replay attack detection result according to the fused scores. The accuracy of identifying the replay attack can be improved, and compared with a face identification technology for verifying the identity in severe visual natural environments such as rain and fog, an anti-spoofing technology of voiceprint of dynamic characteristics and strong identification is adopted, so that the method has higher safety and reliability.

Description

technical field [0001] The invention relates to the technical field of voiceprint recognition, in particular to a speaker verification system and a replay attack detection method thereof. Background technique [0002] Voiceprint Recognition (Voiceprint Recognize) is a technology that extracts the speaker's voice characteristics and speech content information, and automatically verifies the speaker's identity. The Automatic Speaker Verification (ASV) system is a biometric verification system that uses voiceprint recognition technology to verify a person's identity. [0003] Replay Attacks, also known as replay attacks and replay attacks, refer to the attacker sending a packet that the destination host has received to achieve the purpose of deceiving the system. It is mainly used in the identity authentication process to destroy the correctness of authentication. [0004] At present, with the rapid development of automatic speaker verification technology, speaker verification...

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): G06F21/32G10L17/02G10L17/04G10L17/18G10L17/22G06N3/04G06N3/08
CPCG06F21/32G10L17/02G10L17/04G10L17/18G10L17/22G06N3/084G06N3/048G06N3/045
Inventor 窦勇强杨茂林杨皓程
Owner 珠海造极声音科技有限公司
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Patsnap Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Patsnap Eureka Blog
Learn More
PatSnap group products

Browse by: Latest US Patents, China's latest patents, Technical Efficacy Thesaurus, Application Domain, Technology Topic, Popular Technical Reports.

© 2024 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com