Voice Anti-spoofing with LightCNN

This project implements and trains a Countermeasure (CM) system to detect spoofing attacks (replays, synthetic speech, etc.) on the Logical Access (LA) partition of the ASVSpoof 2019 Dataset. The core architecture is a LightCNN (LCNN), implemented from scratch in PyTorch following the specifications from papers:

• STC Antispoofing Systems for the ASVspoof2019 Challenge by Speech Technology Center - used as the foundation for the LightCNN architecture implementation
• A Comparative Study on Recent Neural Spoofing Countermeasures for Synthetic Speech Detection - used for the training recipe and data preparation scheme

Getting Started

These instructions will give you a copy of the project up and running on your local machine for development and training purposes.

Installing

A step-by-step series of examples to get a development environment running.

1. Create and activate a Conda environment:

   conda create -n voice-anti-spoofing python=3.8
   conda activate voice-anti-spoofing

2. Clone this repository and navigate into the project directory:

   git clone https://github.com/ilyalychagin/Voice_Anti-spoofing.git
   cd Voice_Anti-spoofing

3. Install the required Python packages:

   pip install -r requirements.txt

4. Download the Dataset:

   • Download the LA partition from the official ASVSpoof 2019 dataset on Kaggle.
   • Place the downloaded data in the root of the project directory.

5. Start the training process:

   python3 train.py

   The script will begin training the LightCNN model on the ASVSpoof 2019 LA training set.

Results

EER: 7% (Equal error rate)

Acknowledgments

• Acknowledgement to the pytorch_project_template for providing a structured project foundation.