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

Voice identification efficiency optimization method based on dynamic pruning beam prediction

A speech recognition and pruning technology, applied in speech recognition, speech analysis, instruments, etc., can solve the problems of difficulty in confidence estimation, insufficient mining of pruning effectiveness, and rarely used alone

Active Publication Date: 2016-08-10
UNIV OF SCI & TECH OF CHINA
View PDF3 Cites 31 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

They estimate the bundle width of histogram pruning, and the effectiveness of histogram pruning itself is lower than that of bundle width pruning, and it is rarely used alone
In addition, based on the idea of ​​dynamically predicting the pruned bundle width based on the acoustic confidence, its confidence is based on the traditional GMM (Gaussian Mixture Model) modeling method, and it is difficult to estimate the confidence very accurately. In addition, the pruned bundle width estimation modeling is more intuitive Simple, parameter adjustment is based on experience, pruning effectiveness mining is not sufficient

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
  • Voice identification efficiency optimization method based on dynamic pruning beam prediction
  • Voice identification efficiency optimization method based on dynamic pruning beam prediction
  • Voice identification efficiency optimization method based on dynamic pruning beam prediction

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0057] The technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only some of the embodiments of the present invention, not all of them. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without making creative efforts belong to the protection scope of the present invention.

[0058] In continuous speech recognition with a large vocabulary, the search space is very large due to the increase of the dictionary size, so the decoding process is a very time-consuming part of the speech recognizer. Although the traditional pruning algorithm can improve the decoding efficiency by reducing the pruning beam width, it often leads to a sharp decline in recognition performance. The existing improved pruning algorithm has improved th...

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 voice identification efficiency optimization method based on dynamic pruning beam prediction. For problems of relatively many redundancy paths existing in a traditional voice decoding pruning algorithm and insufficient pruning validity existing in an improved algorithm in the prior art, dynamic pruning beam prediction ideas based on acoustics characteristics are proposed, and two specific modeling modes and a corresponding parameter estimation training method are proposed, and thereby voice identification decoding efficiency is improved.

Description

technical field [0001] The invention relates to the technical field of speech recognition, in particular to a speech recognition efficiency optimization method based on dynamic pruning bundle width prediction. Background technique [0002] With the accumulation of large-scale speech annotation data, the improvement of graphics processing unit (GPU) computing speed, and the maturity of deep neural network technology, the effect of large-vocabulary continuous speech recognition has been significantly improved in recent years, and it has become an important tool for human-computer interaction. important way. Although the current voice interaction is mainly based on cloud services, the rise of mobile terminals (such as smart phones) and voice interaction without the Internet also require voice recognizers to be able to adapt to mobile terminals with limited hardware computing resources. Improving the operating efficiency of the speech recognizer has clear significance both in s...

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): G10L15/02G10L15/06G10L15/16G10L15/08
CPCG10L15/02G10L15/063G10L15/08G10L15/16G10L2015/085
Inventor 刘俊华凌震华戴礼荣
Owner UNIV OF SCI & TECH OF CHINA
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