Sonic-Cipher: Spoof-Aware Speaker Verification System

Sonic-Cipher is a Python package for robust speaker verification with integrated spoof detection. It combines automatic speaker verification (ASV) with countermeasure (CM) models to identify and reject spoofed audio inputs. Perfect for secure authentication systems using voice biometrics.

Installation and Configuration

1. Install the package

• From Pypi

pip install sonic-cipher

• Build from Scratch

git clone https://github.com/SonicCypher/Speaker-Verification-Application.git
cd Speaker-Verification-Application
pip install build
python -m build
pip install -e .

2. Create a virtual environment and install the dependencies

pip install -r requirements.txt

4. Export the below envrionment variables

DB_NAME=<your_database>
DB_USER=<your_username>
DB_PASSWORD=<your_password>
DB_HOST=<your_database_host>
DB_PORT=<your_database_port>

Note: Currently sonic-cipher supports only Postgres databases

Usage Example

📝 Register a Speaker

from sonic_cipher import register_user

register_user(username, path1, path2, path3)

• In the registration process, register_user accepts paths of 3 audio clips of the user

🔍 Verify a Speaker

from sonic_cipher import predict_verification

is_verified, confidence_score = register_user(username, path_of_test_audio, device="cpu", threshold=0.1994701042959457)

• Required parameters: username, path of the test audio