论文标题
Voxceleb扬声器识别挑战2020的XX205系统
The xx205 System for the VoxCeleb Speaker Recognition Challenge 2020
论文作者
论文摘要
该报告描述了提交到Voxceleb扬声器识别挑战（VoxSRC）2020的第一和第二轨道的系统，该系统在这两个曲目中排名第二。探索了系统管道的三个关键点：（1）研究多个CNN体系结构，包括RESNET，RES2NET和双路径网络（DPN），以提取X-VECTOR，（2）使用复合角度margin SoftMax损失来训练扬声器模型，以及（3）应用得分正常化和系统融合以提高性能。最佳提交系统以VoxSRC-20评估集进行了衡量，在关闭条件轨道1中获得了$ 3.808 \％$的EER，MIDCF为$ 0.1958 $，分别为$ 3.798 \％$ $ $ 3.798 \％$，而MindCF则分别为$ 0.1942 $ 0.1942 $。
This report describes the systems submitted to the first and second tracks of the VoxCeleb Speaker Recognition Challenge (VoxSRC) 2020, which ranked second in both tracks. Three key points of the system pipeline are explored: (1) investigating multiple CNN architectures including ResNet, Res2Net and dual path network (DPN) to extract the x-vectors, (2) using a composite angular margin softmax loss to train the speaker models, and (3) applying score normalization and system fusion to boost the performance. Measured on the VoxSRC-20 Eval set, the best submitted systems achieve an EER of $3.808\%$ and a MinDCF of $0.1958$ in the close-condition track 1, and an EER of $3.798\%$ and a MinDCF of $0.1942$ in the open-condition track 2, respectively.
