Language recognition method and device and language recognition model training method and device
A language recognition and model training technology, applied in speech recognition, speech analysis, instruments, etc., can solve the problem of low accuracy of language recognition
Active Publication Date: 2021-11-30
合肥讯飞数码科技有限公司
View PDF5 Cites 1 Cited by
- Summary
- Abstract
- Description
- Claims
- Application Information
AI Technical Summary
Problems solved by technology
[0004] The present invention provides a language recognition and language recognition model training method and device to solve the defect of low language recognition accuracy in the prior art when the distribution of languages is unbalanced
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 moreImage
Smart Image Click on the blue labels to locate them in the text.
Smart ImageViewing Examples
Examples
Experimental program
Comparison scheme
Effect test
Embodiment Construction
[0053] In order to make the purpose, technical solutions and advantages of the present invention clearer, the technical solutions in the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the present invention. Obviously, the described embodiments are part of the embodiments of the present invention , but not all examples. Based on the embodiments of the present invention, all other embodiments obtained by persons of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.
[0054]At present, the TV language recognition system based on factor analysis is usually used for language recognition. Before using a speech recognition system for speech recognition, the applied model needs to be trained, and appropriate training samples need to be used during training so that the obtained model can accurately recognize languages.
[0055] However, for some special scenari...
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
Login to View More Abstract
The invention provides a language recognition method and device and a language recognition model training method and device. The recognition method comprises steps that to-be-recognized voice data are inputted into a language recognition model, and a language recognition result outputted by the language recognition model is obtained; wherein the language recognition model comprises a language feature extraction layer and a language classification layer, and the language recognition model is obtained by training based on the following processes, first training is conducted on the language feature extraction layer; parameters of the language feature extraction layer after the first training is completed are fixed, second training is conducted on the language classification layer based on the balanced sample set, a trained language recognition model is obtained, and the number scale of sample voice of all languages in the balanced sample set is consistent. According to the language recognition method and device and the language recognition model training method and device provided by the invention, language recognition can be accurately carried out in a scene with unbalanced language distribution.
Description
technical field [0001] The invention relates to the technical field of natural language processing, in particular to a method and device for language recognition and language recognition model training. Background technique [0002] Language recognition means that the machine automatically judges the language type of the voice data according to the input voice data, such as Chinese, English, French or Japanese. [0003] At present, most TV (Total variability) language recognition systems based on factor analysis perform language recognition on speech data to be recognized. However, when this method is applied to the scene where the language data distribution is unbalanced, for example, when the data distribution ratio of the minority language and the data distribution ratio of the majority language are quite different, the language recognition accuracy is low. Contents of the invention [0004] The present invention provides a language recognition and language recognition...
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
Login to View More Patent Type & Authority Applications(China)
IPC IPC(8): G10L15/06G10L15/00
CPCG10L15/063G10L15/005
Inventor 杨军方磊张梦施志强张胡方四安柳林
Owner 合肥讯飞数码科技有限公司