arXivDaily arXiv每日学术速递 周一至周五更新
arXiv周末暂无论文更新,休息一下吧,周末愉快~~
arXiv 2005.11905eess.AS
收藏

基于神经判别分析的深度说话人嵌入

Neural Discriminant Analysis for Deep Speaker Embedding

Lantian Li, Dong Wang, Thomas Fang Zheng

摘要:

概率线性判别分析(ProbabilityLinearDiscriminantAnalysis,PLDA)是开集分类/验证任务中的一种流行工具。然而,PLDA的高斯假设使其无法应用于数据明显非高斯的情况。在本文中,我们提出了一种新的非线性版本的PLDA命名为神经判别分析(NDA)。该模型采用可逆深度神经网络将复杂分布转换为简单高斯分布,从而可以在转换后的空间中轻松建立线性高斯模型。我们在说话人识别任务上测试了这个NDA模型,其中深度说话人向量(x向量)可能是非高斯的。在两个数据集上的实验结果表明,通过处理x向量的非高斯分布,NDA始终优于PLDA。

英文摘要:

Probabilistic Linear Discriminant Analysis (PLDA) is a popular tool in open-set classification/verification tasks. However, the Gaussian assumption underlying PLDA prevents it from being applied to situations where the data is clearly non-Gaussian. In this paper, we present a novel nonlinear version of PLDA named as Neural Discriminant Analysis (NDA). This model employs an invertible deep neural network to transform a complex distribution to a simple Gaussian, so that the linear Gaussian model can be readily established in the transformed space. We tested this NDA model on a speaker recognition task where the deep speaker vectors (x-vectors) are presumably non-Gaussian. Experimental results on two datasets demonstrate that NDA consistently outperforms PLDA, by handling the non-Gaussian distributions of the x-vectors.

机构、总结、中文摘要均由AI分析而来,可能存在错误,请以论文正文为准。

↑