audio_augmentation

This repository contains audio augmentation tools for PyTorch 2.5+, including codec degradation, noise addition, RIR simulation, and spectral transformations.

Installation

Install directly from git:

pip install git+https://github.com/yourusername/audio_augmentation.git

Or if you have a private repository:

pip install git+ssh://git@github.com/yourusername/audio_augmentation.git

For development installation:

git clone https://github.com/yourusername/audio_augmentation.git
cd audio_augmentation
pip install -e .

Usage

import torchaudio
from audio_augmentation import Degrader

# Initialize degrader with config file
degrader = Degrader(cfg_path="path/to/config.yaml")

# Load audio
waveform, sample_rate = torchaudio.load("input.wav")

# Apply degradation
degraded_waveform = degrader(waveform, sample_rate)

# Save result
torchaudio.save("output.wav", degraded_waveform, sample_rate)

See test/example.py for a complete example.

Features

• Codec Degradation: Apply various audio codecs (Opus, AMR, MP3, Speex, etc.)
• Noise Addition: Add background noise with configurable SNR
• RIR Simulation: Simulate room impulse responses
• Spectral Transformations: Apply various audio effects
• Phone Channel Simulation: Simulate telephone channel characteristics

Requirements

• Python >= 3.10
• PyTorch >= 2.5.1
• torchaudio >= 2.5.1
• See requirements.txt for full list