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

End-to-end multi-language continuous speech stream speech content identification method and system

A technology of speech content and speech recognition model, applied in speech recognition, speech analysis, instruments, etc., can solve problems such as the inability to effectively improve the language distinction of multilingual speech recognition systems

Active Publication Date: 2021-07-06
INST OF ACOUSTICS CHINESE ACAD OF SCI +1
View PDF0 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

Existing speech content recognition methods cannot effectively improve the language discrimination of multilingual speech recognition systems

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
  • End-to-end multi-language continuous speech stream speech content identification method and system
  • End-to-end multi-language continuous speech stream speech content identification method and system
  • End-to-end multi-language continuous speech stream speech content identification method and system

Examples

Experimental program
Comparison scheme
Effect test

Embodiment Construction

[0052] The present invention will be further described now in conjunction with accompanying drawing.

[0053] Such as figure 1 As shown, the present invention provides a kind of end-to-end multilingual continuous speech stream speech content recognition method, and the method comprises:

[0054] Input the speech spectrum features to be recognized into the pre-built segment-level language classification model based on deep neural network, and extract the sentence-level language state posterior probability distribution vector V according to the segment-level language classification model l , to obtain the language classification result of the corresponding language type; wherein, the language classification result of the corresponding language type is the sentence-level language state posterior probability distribution vector V l ; The speech spectrum feature to be recognized is the frequency domain representation obtained by performing Fourier transform on the multilingual con...

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 belongs to the technical field of network communication, and particularly relates to an end-to-end multi-language continuous speech stream speech content recognition method. The method comprises the steps of inputting speech spectrum features to be recognized into a pre-constructed segment-level language classification model based on a deep neural network, and extracting a statement-level language state posterior probability distribution vector; and inputting the to-be-recognized speech spectrum feature sequence of each language type and the statement-level language state posterior probability distribution vector into a pre-constructed multi-language speech recognition model, and outputting a speech recognition result of the corresponding language type.

Description

technical field [0001] The invention belongs to the technical field of network communication and voice recognition, and in particular relates to an end-to-end multilingual continuous voice stream voice content recognition method and system. Background technique [0002] Currently, end-to-end recognition frameworks have been widely used in automatic speech recognition tasks. Since the end-to-end framework does not rely on pronunciation dictionaries in the process of building speech recognition systems, it is more flexible in the process of building speech recognition systems for new languages ​​and multilingual speech recognition systems. Not only that, the end-to-end speech recognition model can directly model the mapping relationship between the sequence of acoustic features and the sequence of text modeling units. Compared with the traditional speech recognition system based on acoustic modeling and language modeling, the end-to-end framework unifies the process of acoust...

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
IPC IPC(8): G10L15/08G10L15/00G10L15/06G10L15/16
CPCG10L15/08G10L15/005G10L15/063G10L15/16
Inventor 徐及林格平刘丹阳万辛张鹏远李娅强刘发强颜永红
Owner INST OF ACOUSTICS CHINESE ACAD OF SCI