One-Class Fake Speech Detection Based on Improved Support Vector Data Description

<div>The t-SNE visualization of svdd output. The red dot represents the natural voice, and the green dot represents the synthetic voice. (a) The t-SNE visualization of AESVDD-1 method. (b) The t-SNE visualization of AESVDD-2 method. (c) The UMAP visualization of AESVDD-1 method. (d) The UMAP visualization of AESVDD-2 method. Samples are randomly selected from AVSspoof 2019 LA development and evaluation datasets.</div>

Security and Communication Networks

fig4

Figure 4

Figure 4: One-Class Fake Speech Detection Based on Improved Support Vector Data Description 