Pulse density thresholds of velvet noise used in artificial reverberation

🚧 This project is a work in progress.

🎧 Overview

This project explores the minimum pulse density required for velvet noise to be subjectively perceived as Gaussian white noise when used in artificial reverberation. The study includes:

• Stimulus generation using synthetic RIRs
• Velvet noise with varying pulse densities
• ABX listening test procedure
• Data analysis and threshold estimation

You can find the full paper in the paper/ folder.

🗂️ Project Structure

. ├── data/       # Original and generated audio data
  ├── notebooks/  # Analysis and visualization
  ├── paper/      # Original paper in pdf
  ├── results/    # Plots, CSV and audio files results
  ├── scripts/    # Scripts to generate audio signals and plots
  ├── src/        # Source code to process audio signals
  |  ├── analytics/
  |  ├── audio/
  |  ├── reverb/
  |  └── visualization/
  ├── setup.py
  ├── LICENSE
  ├── pyproject.toml
  └── README.md

⚙️ Requirements

Dependencies are managed using poetry:

poetry install

Activate environment:

eval "$(poetry env activate)"

🚀 Usage

Example to generate a test stimulus using white noise:

python scripts/generate_stimulus.py -f data/raw_drum.wav -n White -v

Example to generate a test stimulus using velvet noise with a pulse density of 300 p/s:

python scripts/generate_stimulus.py -f data/raw_drum.wav -n Velvet -pd 300 -v

📊 Reproducing Results

See notebooks/analysis.ipynb for data analysis and threshold plots.