search

clovaai / aasist

Official PyTorch implementation of "AASIST: Audio Anti-Spoofing using Integrated Spectro-Temporal Graph Attention Networks"
MIT License
171 stars 41 forks source link

How to Use Feature Extraction Methods #13

Open ahmed-sevimli opened 3 months ago

ahmed-sevimli commented 3 months ago

I would like to try using MFCC/LFCC and other feature extraction methods in this project.

@Jungjee @hsss

Jungjee commented 2 months ago

Hi @ahmed-sevimli, I believe you can start by replacing the conv module here. https://github.com/clovaai/aasist/blob/a04c9863f63d44471dde8a6abcb3b082b07cd1d1/models/AASIST.py#L479

This module derives a feature map directly from a raw waveform input. You can replace them with acoustic features such as MFCC.