Unlock instant, AI-driven research and patent intelligence for your innovation.

Training method and device for speech recognition model

A speech recognition model and speech recognition technology, applied in the field of speech recognition research, can solve problems such as low accuracy and achieve the effect of improving accuracy

Pending Publication Date: 2019-01-01
GREE ELECTRIC APPLIANCES INC
View PDF16 Cites 12 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0005] The invention provides a training method and device for a speech recognition model, which solves the problem of low accuracy in speech recognition with personal accents and dialects in speech recognition technology

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
  • Training method and device for speech recognition model
  • Training method and device for speech recognition model
  • Training method and device for speech recognition model

Examples

Experimental program
Comparison scheme
Effect test

Embodiment 1

[0027] The present invention provides a training method for a speech recognition model, the schematic diagram of the steps is as follows figure 1 ,include:

[0028] Step 101, collect any user voice signal input by any user, and determine the user voice recognition model corresponding to the user;

[0029] Optionally, in implementation, collect user voice signals input by any user, and analyze the user voice signals to obtain user voiceprint features;

[0030] Optionally, user voice signals with different semantics input by the same user may be collected, user voice signals with the same semantics input by the same user may be collected multiple times, or user voice signals with different or identical semantics input by different users may be collected;

[0031] The semantic content of the above-mentioned user voice signal can be a word or phrase, or a sentence, and the above-mentioned user can input according to his actual needs;

[0032] According to the above-mentioned use...

Embodiment 2

[0077] In this embodiment, a training device for a speech recognition model is provided, the device diagram is as follows image 3 mentioned, including:

[0078] A voice signal collection unit 301, configured to collect a user voice signal input by any user, and determine a user voice recognition model corresponding to the user;

[0079] Optionally, in a specific implementation manner, any user voice signal input by any user is collected, and the user voice signal is analyzed to obtain the user voiceprint feature;

[0080] According to the voiceprint features of the user, the corresponding user voice recognition model is searched in the voice recognition model library, and the voice recognition model library includes different user voice recognition models corresponding to different user voiceprint features.

[0081] Optionally, it also includes a speech recognition model creation unit, configured to create a corresponding user in the speech recognition model library accordin...

Embodiment 3

[0091] The invention provides a training device for a speech recognition model, comprising:

[0092] The above-mentioned memory is used to store executable programs;

[0093] In a specific embodiment, the above-mentioned memory is also used to store user voice signals, corresponding to voice text and voice recognition model library;

[0094] The above-mentioned processor is configured to implement the speech recognition model training method described in any one of the above-mentioned Embodiment 1 to Embodiment 4 when executing the above-mentioned executable program.

[0095] A computer storage medium, characterized in that the computer storage medium stores a computer program, and when the computer program is executed, the method for training a speech recognition model described above in any one of the first to fourth embodiments is implemented.

[0096] Optionally, in a specific implementation, it also includes a voice collection device, configured to collect the above-ment...

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 training method and device for a speech recognition model. The method includes the following steps that: user speech signals are collected; a user speech recognition model isadopted to perform speech recognition on the user speech signals, speech text output is obtained when the speech recognition is successful, and the user speech signals are uploaded to a server when the speech recognition fails; and the user speech signals and a corresponding speech text issued by the server are adopted to train the corresponding user speech recognition model. With the training method and device for the speech recognition model of the invention adopted, the problem of the low accuracy of the speech recognition of persons with local accents and dialects during speech recognition can be solved.

Description

technical field [0001] The invention relates to the field of voice recognition research, in particular to a training method and device for a voice recognition model. Background technique [0002] With the development of computer technology and artificial intelligence technology, making computing able to listen and speak is an important development direction of human-computer interaction in the future. Among them, voice has become the most promising way of human-computer interaction, and the use of voice technology to compare human-computer interaction Other interaction methods have more advantages; [0003] In the current speech recognition technology, the speech input is based on standard Mandarin input. The current speech recognition technology has a good accuracy rate for speech recognition input in Mandarin, but the accuracy rate for speech recognition with some personal accents and dialects is very low. Low, at present, the recognition accuracy of the above speech need...

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): G10L15/06G10L15/30G10L15/32
CPCG10L15/063G10L15/30G10L15/32G10L2015/0638
Inventor 廖海霖王慧君毛跃辉王现林陶梦春
Owner GREE ELECTRIC APPLIANCES INC