Published July 12, 2024 | Version v3.0.4
Software Open

josebeo2016/BTS-Encoder-ASVspoof: v3.0.4

Authors/Creators

  • 1. ROR icon Soongsil University

Description

BTS-E: Audio Deepfake Detection Using Breathing-Talking-Silence Encoder

This is official implementation of our paper: BTS-E: Audio Deepfake Detection Using Breathing-Talking-Silence Encoder

Code structure:

  • biosegment: The Sound Segmentation Model, which uses 3 simple GMM for each class: Brathing, Talking and Silence.

  • asvspoof2021/LA/Baseline-Rawnet2-bio: The full pipeline of our work.

Reference

Specially thanks to Hemlata Tak and the ASVspoof organizers for publishing the baseline source code which is easy to follow. We use the Rawnet2 Baseline https://github.com/asvspoof-challenge/2021/tree/main/LA/Baseline-RawNet2 for developing my system.

 

Files

josebeo2016/BTS-Encoder-ASVspoof-v3.0.4.zip

Files (287.1 MB)

Name Size Download all
md5:1e946a96a8158d0c81ba77d4f5c085be
287.1 MB Preview Download

Additional details

Related works

Is supplement to
Software: https://github.com/josebeo2016/BTS-Encoder-ASVspoof/tree/v3.0.4 (URL)