基于BIC和G_PLDA的说话人分离技术研究

doi:10.3969/j.issn.0253-2778.2015.04.005

中国科学技术大学学报 ›› 2015, Vol. 45 ›› Issue (4): 286-293.DOI: 10.3969/j.issn.0253-2778.2015.04.005

• 论著 • 上一篇

基于BIC和G_PLDA的说话人分离技术研究

李锐，卓著，李辉

中国科学技术大学电子科学与技术系安徽合肥 230027

收稿日期:2014-11-04 修回日期:2014-12-13 接受日期:2014-12-13 出版日期:2014-12-13 发布日期:2014-12-13
通讯作者: 李辉
作者简介:李锐，男，1991年生，硕士生. 研究方向：语言信息处理. E-mail: lirui05@mail.ustc.edu.cn

The research of speaker diarization based on BIC and G_PLDA

LI Rui,ZHUO Zhu,LI Hui

Department of Electronic Science and Technology, University of Science and Technology of China, Hefei 230027, China

Received:2014-11-04 Revised:2014-12-13 Accepted:2014-12-13 Online:2014-12-13 Published:2014-12-13

摘要/Abstract

摘要： 传统的以贝叶斯信息准则（Bayesian information criterion，BIC）作为相似性度量的说话人分离技术，在短时对话的分离任务中能取得较好的效果，但是随着对话时长的增加，BIC的单高斯模型不足以描述不同说话人数据的分布，且层次聚类（Hierarchical agglomerative clustering, HAC）时，区分相同说话人和不同说话人的门限值难以划定．针对此问题，提出基于短时BIC和长时G_PLDA的融合方法，充分利用BIC在短时聚类的可靠性和G_PLDA在长时段上的优异区分性，在美国国家标准技术局（NIST）08 Summed测试集上的实验表明，该方法将分类错误率（DER）从BIC基线系统的2.34%降到1.54%，性能相对提升34.2%.

关键词: 说话人分离, 贝叶斯信息准则, 高斯概率线性判别分析, 分类错误率

Abstract: The traditional technology for speaker diarization(SD), which exploits the Bayesian information criterion(BIC) as the similarity metric, can obtain good results in the short dialogue task, but with the length of the dialogue increasing , single Gaussian model of BIC is insufficient to describe the information distribution of different speakers. Moveover, it is difficult to delineate the threshold between the same speakers and different speakers when using hierarchical clustering (HAC). To solve this problem, a fusion method between BIC and G_PLDA was proposed, so as to make full use of the reliability of BIC in short-term clustering and the excellent discriminating power of G_PLDA in long utterancs. A set of experiments based on NIST 08 Summed shows that this new fusion method reduces the diariazation error rate (DER) from 2.34% of BIC baseline system to 1.54%, improving performance of speaker diarization by 34.2%.

Key words: speaker diariazation, BIC, G_PLDA, DER

中图分类号:

TN912.34

李锐，卓著，李辉. 基于BIC和G_PLDA的说话人分离技术研究[J]. 中国科学技术大学学报, 2015, 45(4): 286-293.

LI Rui,ZHUO Zhu,LI Hui. The research of speaker diarization based on BIC and G_PLDA[J]. Journal of University of Science and Technology of China, 2015, 45(4): 286-293.

参考文献

［1］
Moattar M H, Homayounpour M M. A review on speaker diarization systems and approaches[J]. Speech Communication, 2012, 54(10): 1065-1103.
[2] Tranter S E, Reynolds D A. An overview of automatic speaker diarization systems[J]. IEEE Transactions on Audio, Speech, and Language Processing, ,2006, 14(5): 1557-1565.
[3] Makino S, Lee T W, Sawada H. Blind Speech Separation[M]. Berlin, Germany: Springer, 2007.
[4] Wang D L, Brown G J. Computational Auditory Scene Analysis: Principles, Algorithms, and Applications[M]. New Jersey, USA: Wiley, 2006.
[5] Chen S S, Gopalakrishnan P S. Speaker, environment and channel change detection and clustering via the Bayesian information criterion[C]// Proceedings of the DARPA Broadcast News Transcription and Understanding Workshop. Morgan Kaufman, 1998: 127-132.
[6] Ben M, Betser M, Bimbot F, et al. Speaker diarization using bottom-up clustering based on a parameter-derived distance between adapted GMMs[C]// Proceedings of the International Conference on Spoken Language Processing. Jeju, Korea: IEEE Press, 2004: 2329-2332.
[7] Dehak N, Kenny P, 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.

[8] Shum S, Dehak N, Chuangsuwanich E, et al. Exploiting Intra-Conversation Variability for Speaker Diarization[C]// Proceedings of the 11th Annual International Speech Communication Association. Florence, Italy: IEEE Press, 2011: 945-948.
[9] Glembek O, Burget L, Matějka P, et al. Simplification and optimization of i-vector extraction[C]// International Conference on Acoustics, Speech and Signal Processing. Brno, Czech: IEEE Press, 2011: 4516-4519.
[10] Prince S J D, Elder J E. Probabilistic linear discriminant analysis for inferences about identity[C]// 11th International Conference on Computer Vision. Rio de Janeiro, Brazil: IEEE Press, 2007: 1-8.
[11] Kenny P. Bayesian speaker verification with heavy-tailed priors[C]// Proceedings of the Odyssey Speeker and Language Recognition Workshop. Brno, Czech Republic: IEEE Press, 2010: 14.
[12] Kenny P, Stafylakis T, Ouellet P, et al. PLDA for speaker verification with utterances of arbitrary duration[C]// International Conference on Acoustics, Speech and Signal Processing. Vancouver, Canada: IEEE Press, 2013: 7649-7653.
[13] Garcia-Romero D, Espy-Wilson Y. Analysis of I-vector length normalization in speaker recognition systems[C]// Proceedings of the 11th Annual International Speech Communication Association. Florence, Italy: IEEE Press, 2011: 249-252.

()
()

基于BIC和G_PLDA的说话人分离技术研究

The research of speaker diarization based on BIC and G_PLDA

PDF (PC)

可视化

摘要/Abstract

引用本文

使用本文

参考文献

相关文章 0

编辑推荐

Metrics

本文评价