Training method for language recognition model and language recognition method

A language recognition and training method technology, applied in speech recognition, speech analysis, instruments, etc., can solve the problems of long running time and high computational complexity of decoding phoneme sequences

Active Publication Date: 2016-01-27
INST OF ACOUSTICS CHINESE ACAD OF SCI +1
View PDF3 Cites 11 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

[0007] The purpose of the present invention is to overcome the defect that the traditional method based on acoustic spectrum features does not contain speech pronunciation information, overcome the defects of high computational complexity and long running time in decoding phoneme sequences based on the traditional method based on phoneme features, thereby providing a method that reduces computational complexity and improves Language Recognition Method for Recognition Performance

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 for language recognition model and language recognition method
  • Training method for language recognition model and language recognition method
  • Training method for language recognition model and language recognition method

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0045] Now in conjunction with accompanying drawing, the present invention is described in further detail;

[0046] refer to figure 1 , the process of a training method for a language recognition model includes:

[0047] Step 1-1), collecting a certain amount of target language speech data as training data, extracting the phoneme posterior probability;

[0048] Collect a certain amount of speech data in the target language as training data, and through traditional front-end processing of speech data, cut off invalid speech such as silence and music, and retain valid speech from the training data; then extract temporal pattern (TRAP) features for the left and right frequency bands respectively ; The frame length of each frame is 25ms, the frame shift is 10ms, and the left and right frequency bands respectively take the features of 15 frames, so the features of each frame include the surrounding 310ms duration. The TRAP features of the left and right frequency bands are sent t...

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 relates to a training method for a language recognition model and a language recognition method. The language recognition method comprises the steps: extracting the phoneme posterior probability of speech data, converting the phoneme posterior probability into a log domain, conducting dimensionality reduction and mean and variance normalization, and then obtaining phoneme associated features; calculating Baum-Welch statistical magnitude by means of the phoneme associated features, and extracting a phoneme variance factor through the Baum-Welch statistical magnitude; modeling the phoneme variance factor, and establishing an SVM model (a language recognition model); and marking the SVM model by a phoneme variance factor of to-be-recognized speech data, conducting mean and variance normalization on the score, performing linear discriminant analysis and Gauss back end normalization on the normalized score to realize score correction, and finally obtaining a recognition result. Compared with a conventional language recognition method, the language recognition method of the invention has the advantages that the calculation complexity is reduced, the language recognition performance is obviously improved, and the method is highly practical.

Description

technical field [0001] The invention relates to a method for recognizing language information of voice data, and more specifically, the invention relates to a method for recognizing a language based on phoneme-related features. Background technique [0002] With the globalization of information in modern society, language recognition has become one of the research hotspots of speech recognition technology. The purpose of language recognition technology is to be able to manufacture a machine that imitates human thinking to a certain extent to identify the language of speech, that is, to extract the difference information of each language from the speech signal, and use this as a basis to judge the language it belongs to. The extracted speech signal features directly affect the result of language recognition. [0003] Mainstream language recognition technologies include two types: recognition based on acoustic spectrum features and recognition based on phoneme features. [0...

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): G10L15/06G10L15/26
Inventor 周若华王宪亮颜永红索宏彬
Owner INST OF ACOUSTICS CHINESE ACAD OF SCI
Who we serve
  • R&D Engineer
  • R&D Manager
  • IP Professional
Why Eureka
  • Industry Leading Data Capabilities
  • Powerful AI technology
  • Patent DNA Extraction
Social media
Try Eureka
PatSnap group products