Simplification of I-Vector Extraction for Speaker Identification

作     者：XU Longting YANG Zhen SUN Linhui 

作者机构：Broadband Wireless Communication and Sensor Network Technology Key LabNanjing University of Posts and Telecommunications 

出 版 物：《Chinese Journal of Electronics》 (电子学报(英文))

年 卷 期：2016年第25卷第6期

页      面：1121-1126页

核心收录：

学科分类：0711[理学-系统科学] 07[理学] 

基　　金：supported by the National Natural Science Foundation of China(No.60971129,No.61271335,No.61501251) the Scientific Innovation Research Programs of College Graduate in Jiangsu Province(No.CXZZ13 0488) Key Laboratory of the Ministry of Public Security Smart Speech Technology(No.2014ISTKFKT02) the Natural Science Foundation of Jiangsu Province(No.BK20140891) the Natural Science Foundation of the Jiangsu Higher Education Institutions of China(No.13KJB510020) the Science Foundation of Nanjing University of Posts and Telecommunications(No.NY214191) 

主　　题：Speaker identification Closed-set I-vector Symmetric matrix factorization 

摘      要：The identity vector(i-vector) approach has been the state-of-the-art for text-independent speaker recognition, both identification and verification in recent years. An i-vector is a low-dimensional vector in the socalled total variability space represented with a thin and tall rectangular matrix. This paper introduces a novel algorithm to improve the computational and memory requirements for the application. In our method, the series of symmetric matrices can be represented by diagonal expression,sharing the same dictionary, which to some extent is analogous to eigen decomposition, and we name this algorithm Eigen decomposition like factorization(EDLF). Similar algorithms are listed for comparison, in the same condition,our method shows no disadvantages in identification accuracy.