Welcome

I am Xin Wang, a postdoc researcher at Yamagishi Lab, National Institute of Informatics (NII), Japan. I learned a lot from my supervisors at NII and USTC and many colleagues around the world.

My research interests include

• Speech information processing

• Speech synthesis

• Speech deepfake detection & anonymization

See more in my Resume.

     

Messages:

2025.04

Please join us at the ASRU 2025 special session on Frontiers in Deepfake Voice Detection and Beyond

Work done with Xiaoxiao is accepted by IEEE Trans. Information Forensics and Security: A benchmark for multi-speaker anonymization

Work done with Wanying is accepted by ICLR2025 workshop on Watermark: Proactive Detection of Speaker Identity Manipulation with Neural Watermarking

2024.12.20

ASVspoof 5 database has been released. Check Download link on zenodo

2024.08.31

ASVspoof 5 workshop was held at Kos island, Greece. Check the proceeding on ISCA

2024.04.18

Slides for ICASSP paper: Can Large-Scale Vocoded Spoofed Data Improve Speech Spoofing Countermeasure with a Self-Supervised Front End?

2023.11.20

Slides for high-level summaries of DSP+DNN vocoders and another one on anti-spoofing using spoofed data.

2023.10.01

My JST PRESTO project official starts. It is about speech anti-spoofing, privacy, and verification. See more in this post. My title is also promoted to Project Associate Professor.

2023.08.20

Interspeech 2023 tutorial on anti-spoofing has been uploaded to Github.

2023.03.06

Slides for SPSC webinar presentation on “Using vocoders to create spoofed training data for speech spoofing countermeasures” can be found in Slide page. This is a presentation for ICASSP 2023 paper. The code, pre-trained models, and other resources are available at this repo.

2022.09.30

Jupyter notebook of the tutorials on speaker anonymization is available at Slide page.

2022.05.27

The materials for ICASSP short course on neural vocoders are available on Google colab. The old contents are re-edited, and new contents are available (including NSF-HiFiGAN).

2022.01.04

Slides for JST Science Agora talk on speech spoofing detection is available: OCT-2021. There are many spoofing audio samples under various conditions. You can try to detect which one is spoofed.

2021.12.04

The slides for the talk on speech anti-spoofing and anonymization are uploaded. Two Speech Security Issues after Speech Synthesis Boom, 2021 Dec, a talk given at CCF. DEC-2021

2021.09.01

The ASVspoof 2021 Workshop, an official Interspeech 2021 satellite event, will be held online in the form of a Zoom Webinar on September 16th, 2021.

The tentative technical program is available from the ASVspoof website

The registration link is here

2021.07.30

Tutorials on neural vocoders for SPCC 2021 have been uploade online

1. Slides are available on Slideshare ;

2. Hands-on materials on some vocoders (WaveNet, WaveGlow, Blow, NSF) are added to git repo. Please check the tutorials folder.

2021.04.01

Pytorch project is updated https://github.com/nii-yamagishilab/project-NN-Pytorch-scripts:

1. Accumulated commits over the last year;

2. Projects on ASVspoof: many pre-trained models and training recipes;

2020.10.27

New slides have been uploaded:

1. Interspeech 2020 presentation for Cyclic-noise NSF: Conference presentation

2. Odyssey tutorial presentation: Conference presentation

2020.08.30

Repositories from my personal git account have been moved to https://github.com/nii-yamagishilab, including:

1. CURRENNT and related source codes: https://github.com/nii-yamagishilab/project-CURRENNT-public

2. CURRENNT scripts for model training: https://github.com/nii-yamagishilab/project-CURRENNT-scripts

2020.08.20

Tutorial materials on Pytorch NSF models are uploaded. They are based on Jupyter notebooks and can be run on laptop with CPU;

2020.06.31

Pytorch re-implementation of NSF models is available: https://github.com/nii-yamagishilab/project-NN-Pytorch-scripts