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

A Method of Improving the Accuracy of Speech Recognition Based on the Number of Dynamic HMM Events

A technology of the number of events and the accuracy rate, applied in speech recognition, speech analysis, instruments, etc., can solve the problem of the decline of the recognition accuracy rate, and achieve the effect of improving the accuracy rate, improving the recognition accuracy rate, and improving the recognition accuracy rate.

Inactive Publication Date: 2018-06-08
HARBIN INST OF TECH SHENZHEN GRADUATE SCHOOL
View PDF3 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0007] In order to solve the problems existing in the prior art, the present invention proposes a method for improving the accuracy of speech recognition of large-scale isolated words by dynamically changing the number of events of the HMM model, and solves the problem that the recognition accuracy increases with the increase in the number of isolated words. rate drop problem

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
  • A Method of Improving the Accuracy of Speech Recognition Based on the Number of Dynamic HMM Events

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0016] The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments.

[0017] The Hidden Markov HMM model probability parameter that the present invention uses is as follows:

[0018] (1) N, the number of events in the HMM model. The number of events in the HMM model is implicit. In the subsequent expressions, each event in the label model is {S 1 ,S 2 ,...,S N}, the event at time t is q t .

[0019] (2) M, the number of elements in the sequence that can be observed under each event in the HMM model, that is, the number of observed symbols. Mark each observation symbol as V={v 1 ,v 2 ,L,v M}, the observation sequence is O={o 1 ,o 2 ,L,o T}, where o t is an observation symbol in the set V, and T is the length of the observation sequence.

[0020] (3) Event transition probability distribution A=[a ij ],in

[0021] a ij =p[q t+1 = S j |q t = S i ] 1≤i≤N, 1≤j≤N.

[0022] (4) Observation sequence p...

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 present invention provides a method for improving recognition accuracy for large-scale isolated word speech recognition, and establishes a hidden Markov model (HMM) parameter adaptive change mechanism for different isolated words, and solves the problem of different isolated words caused by HMM. In the probability model, the number of events is the same but the recognition accuracy and recognition robustness are low. Experimental results show that the method of the present invention can effectively improve the accuracy of speech recognition of large-scale isolated words on the premise of slightly increasing the amount of recognition calculation. When the number of isolated words to be recognized is 5120, the average accuracy rate of multiple recognition increases from 91% to 97.3%; when the number of isolated words to be recognized is 10240, the average accuracy rate of multiple recognition increases from 87% to 96.3%. Compared with the traditional speech recognition based on the static model of statistical probability, the advantage of adopting the method of the present invention is that the parameters of the recognition model can be adaptively adjusted for different users, thereby improving the accuracy of recognition.

Description

technical field [0001] The invention relates to the field of speech recognition of isolated words, in particular to a method for improving the accuracy of speech recognition of large-scale isolated words. Background technique [0002] After extracting the feature parameters of the speech and obtaining the cluster code, it is very inaccurate to simply rely on the Euclidean distance to determine which word cluster in the thesaurus a certain word to be recognized belongs to. The inherent law of speech is a statistical probability model, and the Euclidean distance reflects the distance between the vector and the cluster center vector, so further training is required on the obtained parameters and codebooks to establish a more accurate statistical probability model, so that It can better reflect the reflection of the characteristic parameters on the internal laws of speech. Hidden Markov (HMM) model is a very good mathematical model that reflects the jump probability of events a...

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): G10L15/00G10L15/14
Inventor 刘明王明江
Owner HARBIN INST OF TECH SHENZHEN GRADUATE SCHOOL
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