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

A method and system for generating speech recognition corpus based on tts

A technology of speech recognition and generation system, applied in speech recognition, speech synthesis, speech analysis and other directions, can solve the problems of low work efficiency, increase the workload of the staff, and high cost, and achieve the goal of improving work efficiency, reducing manual recording, and reducing workload. effect of stress

Active Publication Date: 2021-04-02
ANHUI SEMXUM INFORMATION TECH CO LTD
View PDF10 Cites 0 Cited by
  • Summary
  • Abstract
  • Description
  • Claims
  • Application Information

AI Technical Summary

Problems solved by technology

There are many disadvantages in this way, firstly, the work efficiency is low, secondly, the cost is high, and finally, when the corpus is updated frequently, it will greatly increase the workload of the staff

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
  • A method and system for generating speech recognition corpus based on tts
  • A method and system for generating speech recognition corpus based on tts
  • A method and system for generating speech recognition corpus based on tts

Examples

Experimental program
Comparison scheme
Effect test

example 1

[0057]Example 1: Processing of text. The text "Hello everyone", input the text "Hello everyone" into the TTS converter, the TTS converter analyzes the text "Hello everyone", and splits the text into "you", "we", "big", "home" ", "good", each word has a corresponding text label in the library, where "ni3" corresponds to "you", "men2" corresponds to "men", "da4" corresponds to "big", and "jia1" corresponds to "Home", "hao3" corresponds to "good", and the corresponding voices "you", "men", "big", "home" and "good" are extracted from the speech synthesis database through text annotation, and phrases are formed through linguistic analysis , "ni3men2" corresponds to the voice "you", "da4jia1" corresponds to the voice "everyone", and "hao3" corresponds to the voice "good", forming a TTS voice "Hello everyone". Then, the voice "you" is labeled "ni3men2", the voice "everyone" is labeled "da4jia1", and the voice "good" is labeled "hao3". Among them, the TTS voice "Hello everyone" is p...

example 2

[0058] Example 2: Handling of punctuation marks. The text "Hello, everyone.", the processing method of the text part is the same as the above example 1, "," and "." are marked as a pause for a period of time, where the pause time of "." is greater than the pause time of ",", for example: " , "Pause for 0.5 seconds, "." Pause for 1 second.

example 3

[0059] Example 3: Processing of polyphonic characters. For example, in the text "急急", the text of "结" is marked with "zhao2", "zhe0", "zhuo2", and the text mark of the phrase will have "zhao2ji2", "zhe0ji2", "zhuo2ji2", through and polyphonic characters Thesaurus matching, find "zhao2ji2" in the polyphonic word thesaurus, "zhao2ji2" is the pronunciation of "anxious", so the pronunciation of "anxiety" is marked with "zhao2ji2".

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 speech recognition corpus generation method based on TTS. The speech recognition corpus generation method is characterized by comprising the steps that original speech data is imported into a data pool; a TTS converter imports text annotations and TTS speech data into the data pool at the same time; the data pool analyzes and processes the speech data and the text annotations, and corpora are generated; the data pool exports the corpora, the corpora are stored into the corpus, and backup corpora are generated; the corpus separates the speech part from the text annotation part of the backup corpora, the speech part is fed back to the data pool, and the text annotation part is fed back to the TTS converter. Through the speech recognition corpus generation method andsystem based on TTS, generation and update of the corpus do not depend on manual adding of corpora any more, the system can work uninterruptedly, so that the working efficiency is improved, manual recording is reduced, the cost is reduced, and the work stress of workers is greatly reduced.

Description

technical field [0001] The invention belongs to the technical field of intelligent speech, in particular to a method and system for generating a speech recognition corpus based on TTS. Background technique [0002] Language is the most important, most commonly used and most direct way for human beings to communicate information. Speech intelligent recognition technology, computer automatic speech recognition technology, is a major breakthrough in the realization of man-machine dialogue. It has developed very rapidly in recent years, and its application has been gradually promoted. [0003] The recognition degree of speech recognition technology is closely related to the size of its own corpus. During speech recognition, it is necessary to search for the corresponding corpus in its corpus before recognizing the content of the speech. Once the corpus is too small, the corresponding corpus cannot be found during speech recognition, and the content of the speech will naturally ...

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 Patents(China)
IPC IPC(8): G10L15/06G06F40/279G10L13/04
Inventor 虞焰兴徐勇
Owner ANHUI SEMXUM INFORMATION TECH CO LTD