Supercharge Your Innovation With Domain-Expert AI Agents!

An Automatic Speech Recognition Method Against Attack Immunity

An automatic speech recognition and speech technology, applied in speech recognition, speech analysis, instruments, etc., can solve the problems of large-scale commercialization of speech recognition technology and few defense models, and achieve improved recognition accuracy, good application prospects, and improved The effect of the ability to resist malicious attacks

Active Publication Date: 2022-05-31
SHANDONG UNIV
View PDF2 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0004] In recent years, the voice confrontational attacks that have emerged have become more and more aggressive and have more and more types, but the corresponding defense models have rarely appeared. Commercialization has buried certain hidden dangers

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
  • An Automatic Speech Recognition Method Against Attack Immunity
  • An Automatic Speech Recognition Method Against Attack Immunity
  • An Automatic Speech Recognition Method Against Attack Immunity

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0098] B: Generate a speech confrontation sample according to the speech sampling data obtained in step A.

[0101] B1: Generate the first speech adversarial sample according to the C&W method.

[0102] The C&W (Carlini and Wagner Attacks) method is an existing typical targeted speech adversarial sample

[0106] B13: Use the loss function to perform back-propagation training on the random noise δ added to the speech sampling data, and the most

[0107] The loss function is:

[0108]

[0109] The loss function (1) is defined as the loss function used for the C&W method; wherein minimize represents minimization,

[0111] B2: Generate a second speech adversarial sample according to the targeted PGD attack method.

[0113] B21: Define the value of the parameter, the parameter includes the learning rate, the step size of each iteration and the number of iterations.

[0114] B22: Perform gradient descent according to the following formula, and iteratively generate adversarial speech data...

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 an automatic speech recognition method immune to attacks, comprising the following steps: A: selecting a speech data set to obtain speech sampling data; B: generating speech confrontation samples according to the speech sampling data; C: analyzing the audio in the speech data set And voice confrontation samples are sampled according to the set sampling rate to obtain the input of the defense model; D: Construct the DD‑GAN defense model based on the GAN model; E: Determine the loss functions of the G model and the D model respectively, and use the loss function to The defense model DD-GAN is used for training; F: After the speech to be recognized is processed by the defense model DD-GAN, it is then recognized by the speech recognition model. The invention can improve the ability of the speech recognition model to resist malicious attacks.

Description

An Automatic Speech Recognition Approach for Immunity Against Attacks technical field [0001] The present invention relates to the field of speech recognition, and in particular, to a method for automatic speech recognition that is immune to attacks. Background technique [0002] In recent years, with the development of Internet technology, automatic speech recognition technology has become more and more widely used. Speech Recognition As a natural language processing technology, it can convert the input audio signal into text. At present, speech recognition is used in daily life Applications mainly include Apple Siri, smart speakers and other products. With the improvement of technology, the application of speech recognition in the Internet of Things will become more and more popular. Automatic speech recognition technology needs to preprocess the input audio and extract speech features. Therefore, the study of speech adversarial samples is extremely important. Th...

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): G10L15/22G10L15/26G10L25/24G10L15/06
CPCG10L15/22G10L15/26G10L25/24G10L15/063
Inventor 徐东亮姜开勋郗浩宇
Owner SHANDONG UNIV
Features
  • R&D
  • Intellectual Property
  • Life Sciences
  • Materials
  • Tech Scout
Why Patsnap Eureka
  • Unparalleled Data Quality
  • Higher Quality Content
  • 60% Fewer Hallucinations
Social media
Patsnap Eureka Blog
Learn More

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

© 2025 PatSnap. All rights reserved.Legal|Privacy policy|Modern Slavery Act Transparency Statement|Sitemap|About US| Contact US: help@patsnap.com