Home >

news Help

Publication Information


Title
Japanese:Team Takoyaki submission for VoxCeleb Speaker Recognition Challenge 2020 
English:Team Takoyaki submission for VoxCeleb Speaker Recognition Challenge 2020 
Author
Japanese: 石川 佳佑, 宇都 有昭, 岩野 公司, 篠田 浩一.  
English: Keisuke Ishikawa, Kuniaki Uto, Koji Iwano, Koichi Shinoda.  
Language English 
Journal/Book name
Japanese: 
English:Proc.The VoxSRC Workshop 2020 
Volume, Number, Page        
Published date Oct. 30, 2020 
Publisher
Japanese: 
English:Visual Geometry Group 
Conference name
Japanese:The VoxSRC Workshop 2020 
English:The VoxSRC Workshop 2020 
Conference site
Japanese: 
English: 
Official URL http://www.robots.ox.ac.uk/~vgg/data/voxceleb/data_workshop_2020/takoyaki.pdf
 
Abstract This paper describes a system submitted to the VoxCelebSpeaker Recognition Challenge 2020 (VoxSRC2020). Previ-ously, Stafylakis et al. proposed a self-supervised learningmethod for x-vector-based speaker recognition. This methodtries to reconstruct the features of each input utterance fromits phoneme recognition result and the speaker features for thewhole utterance, but speaker features for each phoneme maybe different. In this paper, we propose a speaker recogni-tion method using phoneme-dependent speaker features. Thismethod concatenates the phone group label obtained fromphoneme recognition to the input features of each frame. Thenit generates frame-by-frame speaker features and use them forreconstruction. In the evaluation experiment, it achieved EERof 3.25%.

©2007 Tokyo Institute of Technology All rights reserved.