Sionna-AcoustiX

[arXiv] [Website] [AVR Code] [BibTex]

This is an acoustic impulse response simulation platform based on the Sionna ray tracing engine. This is used by the NeurIPS'24 Paper Acoustic Volume Rendering for Neural Impulse Response Fields.

This repo contains the Sionna ray tracing engine, which is primarily for radio-frequency signal simulations. We modify the ray interactions with the environments to support Sionna for acosutic simulator and build AcoustiX.

We are still working on this repo to make it more user-friendly. Stay tuned!

Updates

• 2025.12.15: We updated the license.
• 2025.03.04: we update the collect_dataset.py. With detailed configuration files

Contents

This repo contains the official implementation for AcoustiX.

• Installation
• Structure
• Usage
• Config
• Citation

Installation

• install the sionna ray tracing package (we have modifyed the source code on sionna to support acoustic simulation)

cd sionna
pip install .

Project structure

AcoustiX/
├── extract_scene/              # tools to extract the scene files from igibson dataset      
├── img_render/                 # tools to extract the img from igibson dataset
├── simu_config/                # simulation config file
│   ├── basic_config.yml        # basic omnidirectional impulse response config        
│   └── directional_config.yml  # directional impulse response config
├── sionna/                     # sionna ray tracing engine
├── acoustic_absorptions.json   # common material configuration files   
├── ir_utils.py                 # ir metric utils
├── simu_utils.py               # simulation utils 
├── pattern.py                  # speaker or microphone gain pattern file
├── collect_dataset.py          # basic file to collect the impulse response dataset
├── README.md              # Project documentation
└── .gitignore             # Git ignore file

Usage

Minimim usage of AcoustiX to simulate impulse response

python3 collect_dataset.py --save-data # store data to a file 
python3 collect_dataset.py # just show the impulse response data

Use iGibson dataset

1. To use iGibson dataset for simulation, please go download the iGibson dataset
2. Extract the scene files in the ./extrac_scene

python3 extract_scene.py # extract the environment ply files
python3 generate_xml.py # generate the AcoustiX simulator compatible files .XML

Simulation config file

We refer the following configuration to simu_config/basic_config.yml

Ray-Tracing Configuration (rt_config)

Entry	Default Value	Description
max_depth	10	Maximum number of reflections for ray tracing.
num_samples	50000	Number of ray samples for Monte Carlo simulation.
los	true	Enable line-of-sight (direct path).
reflection	true	Enable sound reflections.
diffraction	false	Enable diffraction effects.
scattering	true	Enable scattering effects.
scat_prob	0.00001	Probability of scattering occurring.

Impulse Response Configuration (ir_config)

Entry	Default Value	Description
attn	0.001	Attenuation factor for impulse response.
fs	48000	Sampling frequency in Hz.
ir_len	4800	Impulse response length in samples.
speed	343.8	Speed of sound in m/s.
noise	0.001	Noise level in impulse response.
tx_pattern	uniform	Transmission pattern (e.g., uniform, directional).
rx_pattern	uniform	Receiver pattern (e.g., uniform, directional).

Customized environments

To create your own acoustic environment, please go and check the official tutrial about Sionna: Create your own scene using blender

Liscense

AcoustiX is licensed under a MIT License.

Citation

If you find this project to be useful for your research, please consider citing the paper.

@inproceedings{lanacoustic,
  title={Acoustic Volume Rendering for Neural Impulse Response Fields},
  author={Lan, Zitong and Zheng, Chenhao and Zheng, Zhiwei and Zhao, Mingmin},
  booktitle={The Thirty-eighth Annual Conference on Neural Information Processing Systems}
}