|本期目录/Table of Contents|

[1]曾霞霞,徐戈,吴征远.基于MFCC特征组合参数的说话人识别研究[J].集美大学学报(自然科学版),2016,21(4):317-320.
 ZENG Xia -xia,XU Ge,WU Zheng-yuan.Speaker Recognition Research Based on the Combination of MFCC Feature Parameters[J].Journal of Jimei University,2016,21(4):317-320.
点击复制

基于MFCC特征组合参数的说话人识别研究(PDF)
分享到:

《集美大学学报(自然科学版)》[ISSN:1007-7405/CN:35-1186/N]

卷:
第21卷
期数:
2016年第4期
页码:
317-320
栏目:
数理科学与信息工程
出版日期:
2016-07-28

文章信息/Info

Title:
Speaker Recognition Research Based on the Combination of MFCC Feature Parameters
作者:
曾霞霞12徐戈12吴征远12
(1.闽江学院计算机科学系,福建 福州 350121;2.福建省信息处理与智能控制重点实验室,福建 福州 350121)
Author(s):
ZENG Xia -xia12XU Ge12WU Zheng-yuan12
(1.Department of Computer Science,Minjiang University,Fuzhou 350121,China;2.Fujian Provincial Key Laboratory of Information Processing and Intelligent Control,Fuzhou 350121,China)
关键词:
说话人识别Mel频率倒谱系数高斯混合模型特征参数特征向量
Keywords:
speaker recognitionMel frequency cepstral coefficients(MFCC)Gaussian mixture model(GMM)feature parametereigenvector
分类号:
-
DOI:
-
文献标志码:
A
摘要:
为提高说话人识别系统的识别率,提出了一种提取Mel频率倒谱系数(MFCC)与差分特征组合参数的方法:先对传统的MFCC参数进行特征分量归一化处理,提升MFCC系数的噪声鲁棒性;再用高斯混合模型(GMM)构建了说话人识别系统。使用TIMIT语音库进行实验测试,并比较了不同高斯混合数的MFCC特征参数组合对识别率的影响。结果表明:使用改进的MFCC混合参数明显地提高了说话人的识别率。
Abstract:
To improve the recognition rate of speaker recognition system,this paper presents a method to extract the combination parameter of Mel frequency cepstral coefficients(MFCC) and accelerated coefficient.The normalized eigenvector of the traditional MFCC feature parameters,improved noise robustness of MFCC coefficient.The paper constructs a speaker recognition system using Gaussian mixture model(GMM),and conducts on TIMIT corpus,and analyzs recognition ratios under different Gaussian mixed numbers(GMN) and the combination of MFCC feature parameters.Experiments show that feature parameters combined improved MFCC with accelerated coefficient are beneficial to improve the speaker recognition rate.

参考文献/References:

-

相似文献/References:

备注/Memo

备注/Memo:
-
更新日期/Last Update: 2016-09-24