Open AccessSpeaker Identification Based On Ivector And Xvector
Author(s) -
Xinyu Yuan,
Guanyu Li,
Jiao Han,
Di Wang,
Zhi Tiankai
Publication year - 2021
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1827/1/012133
Subject(s) - speaker recognition , computer science , speaker identification , speech recognition , artificial neural network , artificial intelligence , identification (biology) , speaker diarisation , feature extraction , feature (linguistics) , key (lock) , pattern recognition (psychology) , linguistics , philosophy , botany , computer security , biology
As an important branch of AI (Artificial Intelligence), speaker recogniti-on technology has developed rapidly in recent years. At present, speaker recognition technologies based on traditional methods and deep learning methods are very mature and have achieved good results. As a key technology in human-machine voice intera-ction, speaker recognition has changed human life in many ways. However, there is almost no research on speaker recognition in Tibetan. This paper uses Gaussian mix-ture model and deep neural network TDNN as the theoretical basis. On the Tibetan language corpus, the speaker identification on the ivector of GMM-UBM and the speaker identification on the xvector of TDNN are performed respectively, and the experimental results are compared. The experimental system is divided into prepro-cessing, feature extraction, model training and decision.