Fake-Voice-Detection

Introduction

This repository provides a Python implementation of a Conformer model for fake speech detection, trained on the ASVspoof2019 dataset. The Conformer model is a state-of-the-art neural network architecture that combines the strengths of convolutional neural networks (CNNs) and recurrent neural networks (RNNs), making it well-suited for tasks like speech recognition and audio classification.

Prerequisites

Before you can run the code, ensure you have the following installed:

• Python 3.6 or later.
• NumPy.
• SciPy.
• TensorFlow or PyTorch.
• Librosa.
• tqdm.

Dataset

The ASVspoof2019 dataset is required for training and evaluation. You can download it from the official ASVspoof website: http://www.asvspoof.org/index2019.html

Usage

$ Clone the Repository:

$ https://github.com/Ansh420/Fake-Voice-Detection.git

Install Dependencies:

• pip install -r requirements.txt

Prepare the Data:

Create a directory structure to organize your data. For example:

• data/

  ├── train/

  │ ├── genuine/

  │ └── spoof/

  └── test/

Place the ASVspoof2019 dataset files into the appropriate directories.

Model Architecture

The Conformer model used in this project consists of the following components:

• Convolutional Module: Applies convolutional layers to extract local features from the input audio.
• Attention Module: Uses attention mechanisms to focus on important parts of the input sequence.
• Feed-Forward Module: A fully connected layer that transforms the features.
• Positional Encoding: Encodes positional information into the input sequence to help the model capture temporal dependencies.