HUMAN SPEECH STRUCTURE-BASED SPEECH INTERFERENCE NOISE DESIGN METHOD

WIPO Patent Application WO/2024/103485

A1

A human speech structure-based speech interference noise design method, comprising the following steps: (1) acquiring a large amount of speech data comprising different speakers and different speaking content, extracting voiceprint information and then constructing an initial speech data set; (2) for each user, acquiring a small amount of user speech data, extracting voiceprint information and then matching the closest speech data in the initial speech data set; (3) performing data augmentation on the matching speech data; (4) segmenting the augmented speech data by using a phoneme segmentation algorithm to form a vowel data set and a consonant data set; (5) constructing three noise sequences on the basis of the vowel data set and the consonant data set, and superposing the three noise sequences to obtain interference noise; and (6) continuously generating random interference noise and playing back the random interference noise, and injecting continuous interference noise into an audio record to realize continuous interference. According to the present invention, interference noise cannot be removed from speech, so that leakage of privacy information of a user is avoided.

BA ZHONGJIE (CN)
HUANG PENG (CN)
WEI YAO (CN)
CHENG PENG (CN)
LU LI (CN)
LIN FENG (CN)
LIU ZHENGUANG (CN)
REN KUI (CN)

PCT/CN2022/140663

May 23, 2024

December 21, 2022

Click for automatic bibliography generation  

ZJU HANGZHOU GLOBAL SCIENTIFIC AND TECH INNOVATION CENTER (CN)

G10L25/48; G10L17/18; G10L25/03

CN115208507A	2022-10-18
CN101950498A	2011-01-19
CN103945039A	2014-07-23
CN114337850A	2022-04-12
CN115035903A	2022-09-09
JP2011123141A	2011-06-23

HANGZHOU TIANQIN INTELLECTUAL PROPERTY AGENCY CO., LTD. (CN)

View/Download PDF      PDF Help

Previous Patent: VALVE CORE, PRESSURE RELIEF VALVE, FIREPROOF DATA STORAGE INNER CONTAINER AND FIREPROOF DATA STORAGE...

Next Patent: THERMAL FABRIC MATERIAL AND PREPARATION METHOD