Speaker Recognition System Based on weighted feature parameter

被引：10

作者：

Zhu, Li ^{[1
]}

Yang, Qing ^{[1
]}

机构：

[1] Hubei Univ Nationalities, Sch Informat Engn, Enshi, Hubei, Peoples R China

来源：

INTERNATIONAL CONFERENCE ON SOLID STATE DEVICES AND MATERIALS SCIENCE | 2012年 / 25卷

关键词：

speaker recognition; LPCC; weighted; VQ;

D O I：

10.1016/j.phpro.2012.03.270

中图分类号：

T [工业技术];

学科分类号：

08 ;

摘要：

Speaker recognition is a kind of biometrics technology, which is very popular and widely applied. To improve the effectiveness of feature parameter, a weighted feature extraction method is proposed in this paper. It analyses each component of LPCC and finds out an average contribution sequence. Based on the sequence, LPCC is weighed by each dimension to generate a weighted feature parameter. Finally, based on VQ, a speaker recognition system which is characterized by weighted LPCC is carried out under MATLAB Environment. By experiments, speaker recognition system shows the better performance than traditional model. (C) 2012 Published by Elsevier B.V. Selection and/or peer-review under responsibility of Garry Lee

引用

页码：1515 / 1522

页数：8

共 6 条

[1]

Carey Michael J, 1997, P ICASSP 97, V25

[2] CEPSTRAL ANALYSIS TECHNIQUE FOR AUTOMATIC SPEAKER VERIFICATION [J].

FURUI, S .

IEEE TRANSACTIONS ON ACOUSTICS SPEECH AND SIGNAL PROCESSING, 1981, 29 (02) :254-272

[3]

Jiang Taihui, 2004, COMPUTER ENG APPL, P77

[4] Unsupervised speaker recognition based on competition between self-organizing maps [J].

Lapidot, I ;

Guterman, H ;

Cohen, A .

IEEE TRANSACTIONS ON NEURAL NETWORKS, 2002, 13 (04) :877-887

[5]

Li Xiangping, 2008, RES EXPLORATION LAB, V26

[6]

Zhiping Dan, 2007, CHINA 3 GORGES U NAT, V29

← 1 →