Language-irrelevant key word recognition method and system

A language-independent and keyword-independent technology, applied in speech recognition, speech analysis, instruments, etc., can solve problems that affect system performance, high false alarm rate, incorrect keywords, etc., and achieve the effect of improving system performance and filtering accurately and reasonably

Active Publication Date: 2014-02-05
科大讯飞河北科技有限公司
View PDF3 Cites 35 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, in practical applications, due to the complexity of noise and the variability of accents and channels in the voice data to be detected, the keywords retrieved are often not real keywords, that is, the false alarm rate is high, which affects system performance.

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
  • Language-irrelevant key word recognition method and system
  • Language-irrelevant key word recognition method and system
  • Language-irrelevant key word recognition method and system

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0069] In order to enable those skilled in the art to better understand the solutions of the embodiments of the present invention, the embodiments of the present invention will be further described in detail below in conjunction with the drawings and implementations.

[0070] Under the HMM / Filler framework, the HMM training algorithm based on the MLE (Maximum Likelihood Estimation) criterion and efficient decoding algorithms such as Viterbi and WFST (weighted Finite state Transducer, finite state converter) make the statistical model based on keywords The decoding method of the / Filler model has good operability and generalization in practical applications. However, in a real environment, the speech signal to be detected is often affected by various factors such as noise, channel, and regional population, so that the keyword results retrieved by direct decoding often have high false alarms, which affects system performance. In this regard, the existing HMM / Filler systems gener...

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 language-irrelevant key word recognition method and system. The method includes the steps of receiving a voice signal to be detected, decoding the voice signal according to a pre-constructed decoding network to obtain candidate key words, evaluating confidence coefficients of the candidate key words in different modes, fusing the confidence coefficient evaluation results in different modes to obtain the effective confidence coefficients of the candidate key words, and determining the output key words according to the effective confidence coefficients.

Description

technical field [0001] The invention relates to the technical field of voice keyword recognition, in particular to a language-independent keyword recognition method and system. Background technique [0002] Voice keyword recognition refers to judging from a given voice file or data whether the voice data contains a specific keyword, and determining the position information where the keyword appears. The current mainstream speech keyword recognition is mainly based on speech recognition technology. First, the speech recognizer related to the speech language is used to recognize the text content contained in the speech, and then the specific keyword text and the location information where it appears are retrieved from the text content. Wait. In this method, users can define new keywords more conveniently, which has better scalability. However, since the development and training of the speech recognizer needs to build the acoustic model and language model of the corresponding...

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 Applications(China)
IPC IPC(8): G10L15/10G10L15/14
Inventor 刘俊华魏思胡国平胡郁
Owner 科大讯飞河北科技有限公司
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