Online speech recognition method and system based on recurrent neural network language model

A technology of cyclic neural network and speech recognition, which is applied in speech recognition, speech analysis, instruments, etc., can solve the problem that the RNN language model cannot be applied to previous speech recognition, and achieve the effect of improving accuracy and recognition efficiency

Pending Publication Date: 2021-09-10
北京它思智能科技有限公司
View PDF5 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0003] The present invention provides an online speech recognition method and system based on a recurrent neural network language model to solve

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
  • Online speech recognition method and system based on recurrent neural network language model
  • Online speech recognition method and system based on recurrent neural network language model
  • Online speech recognition method and system based on recurrent neural network language model

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0041] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings, and preferred embodiments described herein are intended to illustrate and explain the present invention, and is not intended to limit the invention.

[0042] Example proposes a method and system for online voice recognition neural network based language model cycle of the present invention, as figure 1 with figure 2 As shown, the method includes:

[0043] Sl, for the original audio feature extraction, feature extraction to obtain complete speech audio;

[0044] S2, the audio input to the speech acoustic model scoring to obtain the acoustic model score;

[0045] S3, the speech audio input to the rescoring WFST RNN speech model and decodes and re-scoring, after obtaining the decoded speech audio;

[0046] S4, the decoded speech audio output as a recognition result.

[0047] Wherein the speech audio input to the re-scoring and RNN WFST speech decoding and re-...

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 provides an online speech recognition method and system based on a recurrent neural network language model. The method comprises the following steps: carrying out feature extraction on an original audio so as to obtain a speech audio obtained after the feature extraction; inputting the speech audio into an acoustic model for scoring to obtain an acoustic model score; inputting the re-scored speech audio into a WFST and RNN speech model for decoding and re-scoring to obtain a decoded speech audio; and outputting the decoded speech audio as a recognition result.

Description

Technical field [0001] The present invention proposes a method and system for online voice recognition neural network based language model circulation, belonging to the technical field of speech recognition. Background technique [0002] RNN current language model based mainly used for conventional ASR recognition obtained N-best or lattice re-scoring do (rescore), that is to say requires two stages, the first is done using conventional speech recognition ASR system, a second stage RNN is the use of a language model to obtain identification results of the first phase of the re-do the scoring. Therefore RNN language model is used only for offline speech recognition system can not be used online voice recognition system. Inventive content [0003] The present invention provides a line-based method and system for speech recognition language model recurrent neural network to solve the conventional problems RNN language model can not be applied prior to speech recognition, the techni...

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): G10L15/01G10L15/02G10L15/06G10L25/18G10L25/30
CPCG10L15/01G10L15/02G10L15/063G10L25/18G10L25/30
Inventor 欧智坚刘岩肖吉孙磊
Owner 北京它思智能科技有限公司
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