Speech recognition method and device

A speech recognition and speech information technology, applied in speech recognition, speech analysis, instruments, etc., can solve the problems of low accuracy rate of speech recognition, unclear pronunciation, limited corpus collection channels and collection costs, etc., and achieve a rich number of parameters , Expand the effect of speech recognition ability

Active Publication Date: 2020-05-19
FUJIAN TQ DIGITAL
View PDF13 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

However, in the existing mainstream speech recognition interactive technology, limited by the collection channels and collection costs of corpus, the current mature corpus is only for adult speech recognition, and there are few children's corpus
Compared with the pronunciation of adults, the speed of children's pronunciation changes rapidly, and inaccurate articulation leads to unclear pronunciation and other problems, resulting in the extremely low accuracy of the existing speech recognition Xu Ting for children
Moreover, most of the existing acoustic models are constructed by combining the Gaussian mixture model and the hidden Markov model. According to the existing feature extraction methods, the extracted feature dimensions are less, which cannot describe the speech itself well.

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
  • Speech recognition method and device
  • Speech recognition method and device
  • Speech recognition method and device

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0071] In order to explain in detail the technical content, structural features, achieved goals and effects of the technical solution, the following will be described in detail in conjunction with specific embodiments and accompanying drawings.

[0072] see figure 1 , is a flow chart of the speech recognition method described in an embodiment of the present invention; the method includes the following steps:

[0073] S1. Store the child's voice information with preset sampling parameters and a preset format;

[0074] S2. Extract the 39-dimensional Mel frequency cepstral coefficients of each frame of audio from the stored children's voice information, and use a mixed Gaussian model to describe the 39-dimensional Mel frequency cepstral coefficients of each frame of audio;

[0075] S3, training the mixed Gaussian model of each state and constructing the hidden Markov model of each state;

[0076] S4. Initialize a deep belief network based on the gmm-hmm model;

[0077] S5. Adj...

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

In order to provide a voice recognition technology for extracting the characteristics more comprehensively and describing the audio more concretely to thereby have a stronger recognition capability, the inventor provides a voice recognition method. The voice recognition method comprises the following steps of storing the children voice information with a preset sampling parameter and a preset format; extracting a 39-dimension mel-frequency cepstral coefficient of each frame of audio from the stored children voice information, and describing the 39-dimension mel-frequency cepstral coefficient of each frame of audio by a Gaussian mixture model; training the Gaussian mixture model (GMM) of various states and constructing a hidden markov model (HMM) of various states; initializing a deep belief network by taking a gmm-hmm model as the basis; adjusting the deep belief network by the trained gmm-hmm model to obtain a deep neural network (dnn)-hmm model based on the deep belief network; utilizing the dnn-hmm model based on the deep belief network to carry out the voice recognition. The inventor also provides a voice recognition device of realizing the above voice recognition method.

Description

technical field [0001] The invention relates to the field of voice information processing and software, in particular to a voice recognition method and device. Background technique [0002] With the continuous development of voice interaction technology, voice recognition technology has become more mature and perfect. However, limited by the user group and the difficulty of collecting corpus, there has been no good way to improve the efficiency of children's speech recognition. [0003] In the existing technology, a speech recognition method for children adds vocal tract length normalization technology in the process of extracting the acoustic features of the corpus, so as to improve the recognition efficiency of children's speech, and by superimposing multiple models to Train the acoustic model. However, in the existing mainstream speech recognition interactive technology, limited by the collection channels and collection costs of corpus, the current mature corpus is only...

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 & AuthorityPatents(China)
IPC IPC(8): G10L15/02G10L15/14G10L15/16G10L25/24
CPCG10L15/02G10L15/144G10L15/148G10L15/16G10L25/24
Inventor范思楠齐昕关胤王维兰吴拥民陈宏展刘德建
OwnerFUJIAN TQ DIGITAL