SPEAKER CONFIRMATION METHOD AND DEVICE

WIPO Patent Application WO/2018/176894

A1

A speaker confirmation method and device. The method comprises: obtaining a second voice (S1); converting a first voice obtained in advance and the second voice into a corresponding first sound spectrogram and second sound spectrogram (S2); performing feature extraction on the first sound spectrogram and the second sound spectrogram by using a convolutional neural network to obtain corresponding first features and second features (S3); performing feature extraction on the first features and the second features by using a time delay neural network to obtain corresponding third features and fourth features (S4); and confirming a speaker according to the third features and the fourth features (S5).

WO/2018/213746	HEADSET FOR ACOUSTIC AUTHENTICATION OF A USER
WO/2020/138662	ELECTRONIC DEVICE AND CONTROL METHOD THEREFOR
WO/2020/210186	SYSTEMS AND METHODS FOR END-TO-END ARCHITECTURES FOR VOICE SPOOFING DETECTION

LI LANTIAN (CN)
WANG DONG (CN)
ZHENG FANG (CN)

PCT/CN2017/114293

October 04, 2018

December 01, 2017

Click for automatic bibliography generation  

UNIV TSINGHUA (CN)

G10L17/02

CN107146624A	2017-09-08
CN106326939A	2017-01-11
CN104903954A	2015-09-09
CN102201236A	2011-09-28
EP2897076A1	2015-07-22

HU, QING ET AL: "Speaker recognition algorithm based on convolutional neural networks", NETINFO SECURITY, no. 4, 30 April 2016 (2016-04-30), pages 57 - 59, ISSN: 1671-1112
PEGAH GHAHREMANI: "Acoustic modeling from the signal domain using CNNs", INTERSPEECH 2016, 12 September 2016 (2016-09-12), pages 3434 - 3437, XP055611812, DOI: 10.21437/Interspeech.2016-1495

CN-KNOWHOW INTELLECTUAL PROPERTY AGENT LIMITED (CN)

View/Download PDF      PDF Help

Previous Patent: ENERGY MANAGEMENT CIRCUIT AND ENERGY MANAGEMENT METHOD FOR TRIBOELECTRIC NANOGENERATOR

Next Patent: PESTICIDE COMPOSITION CONTAINING THIOCARBAMATE COMPOUND AND THIDIAZURON