基于ICA的语音信号表征和特征提取方法

山东大学学报(工学版) ›› 2010, Vol. 40 ›› Issue (4): 19-22.

基于ICA的语音信号表征和特征提取方法

董治强1,刘琚1,邹欣2,杜军1

1. 山东大学信息科学与工程学院, 山东济南 250100; 2. 英国伯明翰大学电子电气与计算机学院, 伯明翰 B152tt

收稿日期:2009-11-16 出版日期:2010-08-16 发布日期:2009-11-16
作者简介:董治强（1983-），男，山东乐陵人，硕士研究生，主要研究方向为独立分量分析及其应用.E-mail: zhqdong111@163.com
基金资助:
国家自然科学基金资助项目（60872024）；山东省自然科学基金资助项目（Y2007G04）；高等学校科技创新工程重大项目培育基金资助项目（708059）

Speech signal representation and feature extraction based on ICA

DONG Zhi-qiang1, LIU Ju1, ZOU Xin2, DU Jun1

1. School of Information Science and Engineering, Shandong University, Jinan 250100, China;
2. School of Electronic, Electrical and Computer Engineering, University of Birmingham, Birmingham B152tt, UK

Received:2009-11-16 Online:2010-08-16 Published:2009-11-16

摘要/Abstract

关键词: 独立分量分析, 语音特征提, 说话人识别

Abstract:

A mathematical derivation was presented to demonstrate the effectiveness of the independent component analysis (ICA) based on signal representation for non-Gaussian speech extraction when corrupted by Gaussian noises. In experiments, the features based on ICA were applied to the speaker identification task. Experimental results showed that the performance of features based on ICA could get obvious improvements compared with the traditional mel-frequency cepstral coefficients (MFCC) features obtained by discrete cosine transform (DCT).

Key words: independent component analysis, speech feature extraction, speaker identification

董治强1,刘琚1,邹欣2,杜军1. 基于ICA的语音信号表征和特征提取方法[J]. 山东大学学报(工学版), 2010, 40(4): 19-22.

DONG Zhi-qiang1, LIU Ju1, ZOU Xin2, DU Jun1. Speech signal representation and feature extraction based on ICA[J]. JOURNAL OF SHANDONG UNIVERSITY (ENGINEERING SCIENCE), 2010, 40(4): 19-22.

参考文献

相关文章 4

[1]	赵洪国,张焕水,张承慧 . 基于主独立内容特征的人脸图像检索方法研究[J]. 山东大学学报(工学版), 2007, 37(4): 0-0 .
[2]	牛新生,叶华,王亮 . 基于二维ICA变换的语音特征提取[J]. 山东大学学报(工学版), 2007, 37(4): 0-0 .
[3]	孙国霞,孙兴华,白树忠,刘琚,孙建德 . 基于主独立内容特征的人脸图像检索方法[J]. 山东大学学报(工学版), 2007, 37(4): 81-84 .
[4]	邹欣,李万龙,刘琚,Peter Jancovic . 基于二维ICA基于二维ICA变换的语音特征提取[J]. 山东大学学报(工学版), 2007, 37(4): 85-88 .

多维度评价

Viewed

Full text

Abstract

Cited

Shared

Discussed