基于深层信息散度最大化的说话人确认方法

doi:10.11959/j.issn.1000-436x.2021133

通信学报 ›› 2021, Vol. 42 ›› Issue (7): 231-237.doi: 10.11959/j.issn.1000-436x.2021133

• 学术通信 • 上一篇

基于深层信息散度最大化的说话人确认方法

陈晨¹^,², 肜娅峰¹, 季超群¹, 陈德运¹^,², 何勇军¹

¹ 哈尔滨理工大学计算机科学与技术学院，黑龙江哈尔滨 150080
² 哈尔滨理工大学计算机科学与技术博士后流动站，黑龙江哈尔滨 150080

修回日期:2021-06-15 出版日期:2021-07-25 发布日期:2021-07-01
作者简介:陈晨（1990− ），女，黑龙江哈尔滨人，博士，哈尔滨理工大学讲师、硕士生导师，主要研究方向为语音信号处理、音频信息分析、说话人识别等
肜娅峰（1997− ），女，河南南阳人，哈尔滨理工大学硕士生，主要研究方向为说话人识别、语音信号处理等
季超群（1995− ），男，黑龙江绥化人，哈尔滨理工大学硕士生，主要研究方向为说话人识别、语音信号处理等
陈德运（1962− ），男，黑龙江哈尔滨人，博士，哈尔滨理工大学教授、博士生导师，主要研究方向为模式识别、机器学习等
何勇军（1980− ），男，四川南充人，博士，哈尔滨理工大学教授、博士生导师，主要研究方向为语音信号处理、图像处理等
基金资助:
国家自然科学基金资助项目(61673142);黑龙江省自然科学基金资助项目(JJ2019JQ0013);黑龙江省博士后专项基金资助项目(LBH-Z20020);黑龙江省普通高校基本科研业务费专项资金资助项目(2020-KYYWF-0341)

Speaker verification method based on deep information divergence maximization

Chen CHEN¹^,², Yafeng RONG¹, Chaoqun JI¹, Deyun CHEN¹^,², Yongjun HE¹

¹ School of Computer Science and Technology, Harbin University of Science and Technology, Harbin 150080, China
² Postdoctoral Research Station of Computer Science and Technology, Harbin University of Science and Technology, Harbin 150080, China

Revised:2021-06-15 Online:2021-07-25 Published:2021-07-01
Supported by:
The National Natural Science Foundation of China(61673142);The Natural Science Foundation of Heilongjiang Province(JJ2019JQ0013);Heilongjiang Postdoctoral Fund(LBH-Z20020);The Fundamental Research Founds for the Central Universities of Heilongjiang Province(2020-KYYWF-0341)

摘要/Abstract

摘要：

针对说话人确认中无法准确捕获特征间非线性关系的问题，提出了一种基于深层信息散度最大化的目标函数表示方法。该方法能通过计算特征所在分布之间相似度，来对特征间的非线性关系进行隐性表示，并在最大化这种统计相关性的优化目标指导下，使深度神经网络向着同类数据更紧凑、异类数据更分散的方向优化，最终达到提升深层特征空间区分性的目标。实验结果表明，相对于其他深度学习方法，所提方法的相对等错误率（EER）最多降低了15.80%，显著提升了系统性能。

关键词: 说话人确认, 目标函数, 深层信息散度, 特征表示学习

Abstract:

To solve the problem that the nonlinear relationship between speaker representations cannot be accurately captured in speaker verification, an objective function based on depth information divergence maximization was proposed.It could implicitly represent the nonlinear relationship between speaker representations by calculating the similarity between their distributions.Under the supervision of the optimization goal of maximizing the statistical correlation, the deep neural network was optimized towards the direction that the within-class data was more compact and the between-class data were far away from each other, and finally the discrimination of deep speaker representation space could be effectively improved.Experimental results show that compared with other deep learning methods, the relative EER of the proposed method is reduced by 15.80% at most, which significantly improves the system performance.

Key words: speaker verification, objective function, deep information divergence, representation learning

中图分类号:

TP391.4

陈晨, 肜娅峰, 季超群, 陈德运, 何勇军. 基于深层信息散度最大化的说话人确认方法[J]. 通信学报, 2021, 42(7): 231-237.

Chen CHEN, Yafeng RONG, Chaoqun JI, Deyun CHEN, Yongjun HE. Speaker verification method based on deep information divergence maximization[J]. Journal on Communications, 2021, 42(7): 231-237.

图/表 4

参考文献 29

[1]	郑方, 李蓝天, 张慧 ,等. 声纹识别技术及其应用现状[J]. 信息安全研究, 2016,2(1): 44-57.
	ZHENG F , LI L T , ZHANG H ,et al. Overview of voiceprint recognition technology and applications[J]. Journal of Information Security Research, 2016,2(1): 44-57.
[2]	张钹, 朱军, 苏航 . 迈向第三代人工智能[J]. 中国科学:信息科学, 2020,50(9): 1281-1302.
	ZHANG B , ZHU J , SU H . Toward the third generation of artificial intelligence[J]. Scientia Sinica (Informationis), 2020,50(9): 1281-1302.
[3]	DEHAK N , KENNY P J , DEHAK R ,et al. Front-end factor analysis for speaker verification[J]. IEEE Transactions on Audio,Speech,and Language Processing, 2011,19(4): 788-798.
[4]	VESTMAN V , KINNUNEN T . Supervector compression strategies to speed up I-vector system development[C]// Odyssey 2018 The Speaker and Language Recognition Workshop.[S.n.:s.l.], 2018: 357-364.
[5]	MA J B , SETHU V , AMBIKAIRAJAH E ,et al. Generalized variability model for speaker verification[J]. IEEE Signal Processing Letters, 2018,25(12): 1775-1779.
[6]	CHEN C , HAN J Q . TDMF:task-driven multilevel framework for end-to-end speaker verification[C]// 2020 IEEE International Confe rence on Acoustics,Speech and Signal Processing. Piscataway:IEEE Press, 2020: 6809-6813.
[7]	高荣春, 韩纪庆, 张磊 . 说话人识别中基于最大后验概率的通道补偿方法[J]. 通信学报, 2009,30(3): 99-103.
	GAO R C , HAN J Q , ZHANG L . Channel compensation of speaker identification based on maximum a posteriori[J]. Journal on Communications, 2009,30(3): 99-103.
[8]	汪海彬, 郭剑毅, 毛存礼 ,等. 基于通用背景-联合估计(UB-JE)的说话人识别方法[J]. 自动化学报, 2018,44(10): 1888-1895.
	WANG H B , GUO J Y , MAO C L ,et al. Speaker recognition based on universal background-joint estimation(UB-JE)[J]. Acta Automatica Sinica, 2018,44(10): 1888-1895.
[9]	VARIANI E , LEI X , MCDERMOTT E ,et al. Deep neural networks for small footprint text-dependent speaker verification[C]// 2014 IEEE International Conference on Acoustics,Speech and Signal Processing. Piscataway:IEEE Press, 2014: 4052-4056.
[10]	SNYDER D , GARCIA-ROMERO D ,, POVEY D , et al . Deep neural network embeddings for text-independent speaker verification[C]// Interspeech 2017. Piscataway:IEEE Press, 2017: 999-1003.
[11]	SNYDER D , GARCIA-ROMERO D ,, SELL G , et al . X-vectors:robust DNN embeddings for speaker recognition[C]// 2018 IEEE International Conference on Acoustics,Speech and Signal Processing. Piscataway:IEEE Press, 2018: 5329-5333.
[12]	LECUN Y , BOSER B , DENKER J S ,et al. Backpropagation applied to handwritten zip code recognition[J]. Neural Computation, 1989,1(4): 541-551.
[13]	VILLALBA J , CHEN N , SNYDER D ,et al. State-of-the-art speaker recognition for telephone and video speech[C]// Proceeding of the Twenty Annual Conference of the International Speech Communication Association. Piscataway:IEEE Press, 2019: 1488-1492.
[14]	SNYDER D , GARCIA-ROMERO D ,, SELL G , et al . Speaker recognition for multi-speaker conversations using X-vectors[C]// 2019 IEEE International Conference on Acoustics,Speech and Signal Processing. Piscataway:IEEE Press, 2019: 5796-5800.
[15]	ZHANG R T , WEI J G , LU W H ,et al. ARET:aggregated residual extended time-delay neural networks for speaker verification[C]// Interspeech 2020. Piscataway:IEEE Press, 2020: 946-950.
[16]	YU Y Q , LI W J . Densely connected time delay neural network for speaker verification[C]// Interspeech 2020. Piscataway:IEEE Press, 2020: 921-925.
[17]	NAGRANI A , CHUNG J S , ZISSERMAN A . VoxCeleb:a large-scale speaker identification dataset[C]// Interspeech 2017. Piscataway:IEEE Press, 2017: 2616-2620.
[18]	BHATTACHARYA G , ALAM M J , GUPTA V ,et al. Deeply fused speaker embeddings for text-independent speaker verification[C]// Interspeech 2018. Piscataway:IEEE Press, 2018: 3588-3592.
[19]	ZHANG C L , KOISHIDA K , HANSEN J H L . Text-independent speaker verification based on triplet convolutional neural network embeddings[J]. IEEE/ACM Transactions on Audio,Speech,and Language Processing, 2018,26(9): 1633-1644.
[20]	陈莹, 陈湟康 . 基于多模态生成对抗网络和三元组损失的说话人识别[J]. 电子与信息学报, 2020,42(2): 379-385.
	CHEN Y , CHEN H K . Speaker recognition based on multimodal generative adversarial nets with triplet-loss[J]. Journal of Electronics ＆Information Technology, 2020,42(2): 379-385.
[21]	HUANG Z L , WANG S , YU K . Angular softmax for short-duration text-independent speaker verification[C]// Interspeech 2018. Piscataway:IEEE Press, 2018: 3623-3627.
[22]	NOVOSELOV S , SHULIPA A , KREMNEV I ,et al. On deep speaker embeddings for text-independent speaker recognition[C]// Odyssey 2018 The Speaker and Language Recognition Workshop. Piscataway:IEEE Press, 2018: 378-385.
[23]	YU Y Q , FAN L , LI W J . Ensemble additive margin softmax for speaker verification[C]// 2019 IEEE International Conference on Acoustics,Speech and Signal Processing. Piscataway:IEEE Press, 2019: 6046-6050.
[24]	WEI Y H , DU J Z , LIU H . Angular margin centroid loss for text-independent speaker recognition[C]// Interspeech 2020. Piscataway:IEEE Press, 2020: 3820-3824.
[25]	KULLBACK S , LEIBLER R A . On information and sufficiency[J]. The Annals of Mathematical Statistics, 1951,22(1): 79-86.
[26]	BELGHAZI M.I , BARATIN A , RAJESHWAR S ,et al. Mutual information neural estimation[C]// Proceeding of the Thirty-Fifth International Conference on Machine Learning. Piscataway:IEEE Press, 2018: 531-540.
[27]	REYNOLDS D A , QUATIERI T F , DUNN R B . Speaker verification using adapted Gaussian mixture models[J]. Digital Signal Processing, 2000,10(1/2/3): 19-41.
[28]	龙华, 杨明亮, 邵玉斌 . 基于特征流融合的带噪语音检测算法[J]. 通信学报, 2020,41(4): 134-142.
	LONG H , YANG M L , SHAO Y B . Noisy voice detection algorithm based on feature stream fusion[J]. Journal on Communications, 2020,41(4): 134-142.
[29]	MAATEN L , HINTON G . Visualizing data using t-SNE[J]. Journal of Machine Learning Research, 2008,9(11): 2579-2605.

方法	EER	Min DCF
GMM-UBM	15.00%	0.80
I-vector+PLDA	8.80%	0.73
Siamese VGG-M	7.85%	0.68
Triplet VGG-M	7.71%	0.66
AM-softmax VGG-M	7.35%	0.63
ID-max VGG-M	6.61%	0.61

基于深层信息散度最大化的说话人确认方法

Speaker verification method based on deep information divergence maximization

在线阅读

PDF下载

可视化

摘要/Abstract

引用本文

使用本文

图/表 4

参考文献 29

相关文章 1

Metrics

推荐阅读 0