Centroid-aware local discriminative metric learning in speaker verification

Centroid-aware local discriminative metric learning in speaker verification
复制标题

说话人验证中的质心感知局部判别度量学习

DOI:
10.1016/j.patcog.2017.07.007
复制
发表时间:
2017-12
影响因子:
8
通讯作者:
Hu Baogang
Hu Baogang
中科院分区:
计算机科学1区
文献类型:
--
作者:
Sheng Kekai;Dong Weiming;Li Wei;Razik Joseph;Huang Feiyue;Hu Baogang

文献摘要

参考文献

被引文献

相似文献

我们提出了一种新的机制,为自动说话人确认(ASV)中针对类不平衡的有效学习铺平了道路,并改进了身份向量(I-向量)的表示。这种洞察力是为了有效地利用ASV内部的内在结构--语体中心先验。具体而言:(1)为了保证学习效率不受类别不平衡的影响,提出了质心感知的平衡升压抽样来收集平衡的小批次;(2)为了加强对小批次的局部判别建模,在ASV的特定修改中采用了邻域成分分析(NCA)和磁损耗(MNL)。该集成创建了自适应NCA(AdaNCA)和线性MNL(LMNL)。数值结果表明,LMNL是一种具有竞争力的I向量低维投影(SRE2008的EER=3.84%,SRE2010的EER=1.81%),与线性判别分析(LDA)相比具有竞争优势。AdaNCA(SRE2008的EER=4.03%,SRE2010的EER=2.05%)也表现良好。此外,为了便于以后对增强采样的研究,建立了增强采样、铰链损失和数据增强之间的联系,这有助于进一步理解增强采样的行为。
We propose a new mechanism to pave the way for efficient learning against class-imbalance and improve representation of identity vector (i-vector) in automatic speaker verification (ASV). The insight is to effectively exploit the inherent structure within ASV corpus — centroid priori. In particular: (1) to ensure learning efficiency against class-imbalance, the centroid-aware balanced boosting sampling is proposed to collect balanced mini-batch; (2) to strengthen local discriminative modeling on the mini-batches, neighborhood component analysis (NCA) and magnet loss (MNL) are adopted in ASV-specific modifications. The integration creates adaptive NCA (AdaNCA) and linear MNL (LMNL). Numerical results show that LMNL is a competitive candidate for low-dimensional projection on i-vector (EER=3.84% on SRE2008, EER=1.81% on SRE2010), enjoying competitive edge over linear discriminant analysis (LDA). AdaNCA (EER=4.03% on SRE2008, EER=2.05% on SRE2010) also performs well. Furthermore, to facilitate the future study on boosting sampling, connections between boosting sampling, hinge loss and data augmentation have been established, which help understand the behavior of boosting sampling further.
DOI: --
发表时间: 2010
影响因子: --
作者:
Mohammed Senoussaoui;P. Kenny;N. Dehak;P. Dumouchel
通讯作者: Mohammed Senoussaoui;P. Kenny;N. Dehak;P. Dumouchel
DOI: 10.1109/cvpr.2016.554
发表时间: 2016-06
期刊: 2016 IEEE Conference on Computer Vision and Pattern Recognition (CVPR)
影响因子: --
作者:
Olivier Canévet;F. Fleuret
通讯作者: Olivier Canévet;F. Fleuret
DOI: 10.1109/icassp.2011.5947363
发表时间: 2011-05
期刊: 2011 IEEE International Conference on Acoustics, Speech and Signal Processing (ICASSP)
影响因子: --
作者:
N. Dehak;Z. Karam;D. Reynolds;Réda Dehak;W. Campbell;James R. Glass
通讯作者: N. Dehak;Z. Karam;D. Reynolds;Réda Dehak;W. Campbell;James R. Glass
DOI: 10.1007/978-3-642-83520-9_60
发表时间: 1988
期刊: --
影响因子: --
作者:
N. Lack
通讯作者: N. Lack
DOI: 10.1007/978-3-642-35289-8
发表时间: 2012-11
期刊: --
影响因子: --
作者:
Grgoire Montavon;Genevive Orr;Klaus-robert Mller
通讯作者: Grgoire Montavon;Genevive Orr;Klaus-robert Mller