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

Command word recognition self-adaptive optimization method for carrying out feedback on user speech recognition result

A technology for speech recognition and recognition results, applied in speech recognition, speech analysis, instruments, etc., can solve problems such as the decline of recognition effect and recognition rejection effect, and the inability to provide a more accurate network, so as to improve the recognition and recognition rejection effect, The effect of improving the recognition efficiency

Inactive Publication Date: 2014-01-22
IFLYTEK CO LTD
View PDF4 Cites 24 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

In this way, the identification network is fixed, and a more accurate network cannot be given according to the usage habits of different users.
Especially when the number of command words is too large, the recognition effect and rejection effect will be significantly reduced

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
  • Command word recognition self-adaptive optimization method for carrying out feedback on user speech recognition result
  • Command word recognition self-adaptive optimization method for carrying out feedback on user speech recognition result

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0017] The recognition process and the vocabulary grouping self-optimization and adjustment process will be described in detail below in combination with the flow chart.

[0018] Such as figure 1 Shown, the concrete realization steps of the present invention are as follows:

[0019] (1) First, give weight scores to all supported command words (the score is based on experience), each command word will have a weight score, and then divide all command words into two groups according to the score, a group of words is user Commonly used command words, and the other group is users' infrequently used command words;

[0020] (2) Then the two sets of command words are used to build speech recognition networks, namely the commonly used command word network and the uncommon word recognition network: firstly, the common command word set is sent to the recognition system, and then the recognition system will recognize each word in the command word set. command words into corresponding ac...

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 command word recognition self-adaptive optimization method for carrying out feedback on a user speed recognition result. The method comprises the steps that all supported command words are divided into two groups, and the support command words include user frequently-used command words and user infrequently-used command words; a frequently-used command word network and an infrequently-used command word network are respectively established for the two groups of command words; an input speech signal is inputted to the frequently-used command word network to be recognized, meanwhile speech feature data extracted in the recognition process are stored in a cache; if a recognition result reaches a preset confidence coefficient threshold, the recognition result is directly outputted, the weight of the corresponding command word is adjusted according to the feedback recognition result; if not, the data are directly fetched and inputted to the infrequently-used command word recognition network to be recognized; if the data are inputted to the infrequency-used command word recognition network to be recognized, and a recognition result reaches a preset confidence coefficient threshold, the recognition result is fed back to adjust the weight of the corresponding command word; if not, the recognition is refused. According to the command word recognition self-adaptive optimization method for carrying out feedback on the user speed recognition result, command word recognition networks can be automatically optimized according to use habits of different users, and the recognition effect and the recognition refusal effect are improved.

Description

technical field [0001] The invention relates to a speech recognition method, in particular to a command word recognition self-adaptive optimization method for user speech recognition result feedback. Background technique [0002] The command word speech recognition system mainly completes the function of automatic speech recognition of the pre-defined command word set. When using the system, the user needs to first send the defined set of command words into the recognition system, and then the recognition system will convert each command word in the set of command words into a corresponding acoustic phoneme sequence, and then combine these phoneme sequences into A graph network, which we call a recognition network. [0003] With the built command word recognition network, the user only needs to say the voice command in the command word set when using the recognition system, and send it to the recognition system, and the recognition system will recognize the matching words i...

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): G10L15/00
Inventor 张伟梅珂陈盛鹿晓亮单言丰
Owner IFLYTEK CO LTD
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