What is it about?
The Audio Deep Synthesis Detection Challenge (ADD 2022) has been held to address the problem of fake audio detection in situations of diverse background noises and disturbances. We propose a Frequency Feature Masking (FFM) augmentation considering noisy environments. We applied FFM and mixup augmentation on five spectrogram-based deep neural network architectures that performed well for spoofing detection using mel-spectrogram and constant Q transform (CQT) features. Our best submission achieved 23.8\% in EER and ranked 3rd on track 1.
Featured Image
Why is it important?
When developing services using voice, it is necessary to develop methods to prevent attacks that try to exploit it.
Perspectives
In the ADD competition, it seems important to make a model that considers noisy environmental signals included in low and high frequencies. The team that won first place in the competition also considered the low pass filter and the high pass filter and modeled it reflecting this.
Il-Youp Kwak
Chung-Ang University
Read the Original
This page is a summary of: Low-quality Fake Audio Detection through Frequency Feature Masking, October 2022, ACM (Association for Computing Machinery),
DOI: 10.1145/3552466.3556533.
You can read the full text:
Contributors
The following have contributed to this page
