基于二维ICA基于二维ICA变换的语音特征提取

山东大学学报(工学版)

基于二维ICA基于二维ICA变换的语音特征提取

邹欣^1,2, 李万龙¹，刘琚¹, Peter Jancovic²

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

收稿日期:2006-12-21 修回日期:1900-01-01 出版日期:2007-08-24 发布日期:2007-08-24
通讯作者: 邹欣

Speech feature extraction based on 2-D independent component analysis

ZOU Xin^1,2，LI Wan-long¹,LIU Ju¹,Peter Jancovic²

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

Received:2006-12-21 Revised:1900-01-01 Online:2007-08-24 Published:2007-08-24
Contact: ZOU Xin

摘要/Abstract

摘要： 独立成分分析 (ICA)方法已经被广泛地应用于语音信号处理中. 讨论了ICA方法在语音信号特征提取中的应用.ICA被应用在对数Mel滤波器组变换域中来代替常用的离散余弦变换，后者被应用来得到Mel倒谱系数(MFCC)特征.我们将应用一种新的方法即二维ICA方法来发掘语音信号的时域跟频域的信息，从而提高语音特征的效率跟噪声鲁棒性.这些特征被用于基于高斯混合模型的说话人识别应用中.仿真结果表明我们得到的时频二维特征优于传统的一维特征.

关键词: 独立成份分析, 语音特征提取, 说话人识别

Abstract: Independent Component Analysis (ICA) has been widely used in speech signal processing performance. The application of ICA to speech feature extraction is investigated. Normally the ICA algorithm is applied in the logarithm Mel-scaled filter bank domain to replace the discrete cosine transform (DCT), which is applied to obtain Mel-Frequency Cepstral Coefficients (MFCCs). A novel 2-D ICA algorithm is proposed, with which one can explore the spectral and temporal information of speech signals, and improve the effectiveness and noise robustness of speech features. The evaluation is presented for GMM-based speaker recognition task. The experimental results show that the proposed 2-D features provide better performance than normal 1-D features.

Key words: ICA, speech feature extraction, speaker identification

中图分类号:

TN912.34

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 .

ZOU Xin,LI Wan-long,LIU Ju,Peter Jancovic . Speech feature extraction based on 2-D independent component analysis[J]. JOURNAL OF SHANDONG UNIVERSITY (ENGINEERING SCIENCE), 2007, 37(4): 85-88 .

参考文献

多维度评价

Viewed

Full text

Abstract

Cited

Shared

Discussed

[1]	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.
[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 .