trungdev/procedural-engine-sounds

1---2pretty_name: Procedural Engine Sounds Dataset3task_categories:4- audio-to-audio5- audio-classification6- automatic-speech-recognition7tags:8- audio9- audio-dataset10- engine-sounds11- combustion-engine12- procedural-generation13- time-aligned14- rpm15- torque16- automotive17- nvh18- vehicle-acoustics19- sound-synthesis20- noise-free21size_categories:22- 10B<n<100B23license: cc-by-nc-4.024---25 26 27# Procedural Engine Sounds Dataset28 29## Dataset Description30 31The Procedural Engine Sounds Dataset is a comprehensive collection of synthetically generated and annotated engine audio samples. This dataset contains procedurally generated high-resolution engine sounds free of confounding noises, with detailed time-aligned annotations, designed for research in audio processing, vehicle acoustics, and synthetic sound generation.32 33## Associated Publication34 35The dataset and generation methodology are described in the following paper:36 37Doerfler, R., & Wyse, L. (2026).  38*Analysis-Driven Procedural Generation of an Engine Sound Dataset with Embedded Control Annotations.*  39https://arxiv.org/abs/2603.0758440 41## Dataset Details42 43### Dataset Summary44 45- **Repository:** procedural-engine-sounds46- **Version:** 1.047- **Publication Year:** 202548- **License:** CC BY-NC 4.049 50### Research Applications51 52- **Audio Generation**: Train models to generate realistic engine sounds conditioned to time-varying engine operation states53- **Audio Classification**: Predict RPM and Engine Torque based on audio signals54- **Audio Analysis**: Research vehicle acoustics and engine sound patterns55- **Sound Synthesis**: Develop procedural audio generation techniques56- **Data Augmentation**: Use as augmentation material for in-cabin speech detection and recognition, noise suppression or other related tasks57 58### Technical Specifications59 60This dataset contains only audio signals - no textual or linguistic content. Both the engine sounds and annotations (RPM/torque information) are provided as audio signals at 48 kHz sample rate in WAV format.61 62## Dataset Structure63 64### Data Organization65 66The dataset is organized into 8 distinct sets with two categories:67 68**Full Sets** (A, B, C, D):69- 3,068 files total (across 4 sets)70- ~9.83 hours of audio total71- ~12.65 GB total72- **Per set (average): ~767 files, ~2.46 hours, ~3.16 GB**73 74**Large Sets** (E, F, G, H):75- 2,867 files total (across 4 sets)76- ~9.18 hours of audio total77- ~11.82 GB total78- **Per set (average): ~717 files, ~2.30 hours, ~2.96 GB**79 80**Dataset Total: 5,935 files, ~19.01 hours, ~24.47 GB**81 82### File Organization83 84```85README.txt (this file)86USAGE.txt (quick start guide)87audio/88├── A_full_set/ (767 files, ~2.46 hours, ~3.16 GB)89├── B_full_set/ (767 files, ~2.46 hours, ~3.16 GB)90├── C_full_set/ (767 files, ~2.46 hours, ~3.16 GB)91├── D_full_set/ (767 files, ~2.46 hours, ~3.16 GB)92├── E_large_set/ (717 files, ~2.30 hours, ~2.96 GB)93├── F_large_set/ (717 files, ~2.30 hours, ~2.96 GB)94├── G_large_set/ (717 files, ~2.30 hours, ~2.96 GB)95└── H_large_set/ (717 files, ~2.30 hours, ~2.96 GB)96metadata/97├── A_full_set_summary.json98├── A_full_set_stats.csv99├── B_full_set_summary.json100├── B_full_set_stats.csv101└── … (16 metadata files total)102```103 104### File Formats105 106**Audio Files:**107- Format: WAV108- Sample Rate: 48 kHz109- Channels: 4 (quad-channel)110- Bit Depth: 16 bit111 112**Metadata Files:**113- Summary files: JSON format114- Statistics files: CSV format (comma-separated values)115 116### Data Structure117 118Each audio file contains **4-channel audio** at 48 kHz sample rate:119- **Channel 1-2**: Stereo engine sound audio120- **Channel 3**: Engine speed (RPM × 0.0001) as continuous audio signal121- **Channel 4**: Engine torque (Nm × 0.001) as continuous audio signal122 123### Metadata Structure124 125#### Summary Files (.json)126Per-set statistics including:127- `num_files`: Number of audio files in set128- `total_duration_*`: Duration in seconds/minutes/hours129- `total_size_gb`: Storage size in GB130- `rpm_distribution`: Statistical distribution (min, max, mean, std, percentiles)131- `torque_distribution`: Statistical distribution (min, max, mean, std, percentiles)132 133#### Statistics Files (.csv)134Per-file metrics with columns:135- `filename`: Audio file name136- `samplerate`: Sample rate (48 kHz)137- `duration_sec`: File duration in seconds138- `size_MB`: File size in megabytes139- `rpm_min/max/mean/std`: RPM statistics for the file140- `torque_min/max/mean/std`: Torque statistics for the file141 142### Data Access143 144Each audio file is a standard WAV file containing a 4-channel audio array at 48 kHz sample rate. When loaded, you receive the raw multichannel audio data from which RPM and torque information can be extracted from channels 3 and 4 respectively.145 146### Audio Signal Encoding147 148- **RPM Signal**: Channel 3 contains engine speed values in RPM scaled by 0.0001 (multiply by 10,000 to get actual RPM)149- **Torque Signal**: Channel 4 contains torque values in Newton meters scaled by 0.001 (multiply by 1,000 to get actual Nm)150- **Engine Audio**: Channels 1-2 contain the stereo procedural engine sound151 152## Technical Requirements153 154To work with this dataset, you will need:155- Audio processing software capable of reading multi-channel WAV files156- Programming languages: Python (recommended with librosa, soundfile, or scipy), MATLAB, R, or similar157- For metadata: JSON and CSV reading capabilities158 159## Dataset Creation160 161### Source Data162 163All audio samples are synthetically generated using procedural audio synthesis techniques. 164No real-world engine recordings were used for audio generation. 165Dataset results were thoroughly analysed and compared to real world recordings to verify representativeness and similarity regarding engine order magnitudes and harmonic deviations.166 167### Annotations168 169Annotations were created during the generation process, with additional manual verification for quality assurance.170 171## Considerations for Using the Data172 173### Social Impact of Dataset174 175This dataset enables research in:176- Automotive audio simulation177- Vehicle sound design178- Audio processing algorithms179- Synthetic data generation techniques180 181### Discussion of Biases182 183As a synthetic dataset, it reflects the biases inherent in the procedural generation algorithms and may not capture all real-world engine sound variations.184 185### Other Known Limitations186 187- Limited to procedurally generated sounds188- May not represent all engine types or acoustic environments189- Generated with empirically determined synthesis parameters, hence contains fictional engine types and exhaust pipe configurations190 191## License and Usage192 193### License194 195This dataset is released under CC BY-NC 4.0 license (Creative Commons Attribution-NonCommercial 4.0 International).196 197**Attribution Required**: Please cite this dataset in any research or publications.198 199### Citation200 201```bibtex202@dataset{doerfler_2025_procedural_engine_sounds,203  author       = {Doerfler, Robin},204  title        = {Procedural Engine Sounds Dataset},205  month        = {August},206  year         = 2025,207  publisher    = {Zenodo},208  version      = {1.0},209  doi          = {10.5281/zenodo.16883336},210  url          = {https://doi.org/10.5281/zenodo.16883336}211}212 213@misc{doerfler2026analysisdrivenproceduralgenerationengine,214      title={Analysis-Driven Procedural Generation of an Engine Sound Dataset with Embedded Control Annotations}, 215      author={Robin Doerfler and Lonce Wyse},216      year={2026},217      eprint={2603.07584},218      archivePrefix={arXiv},219      primaryClass={cs.SD},220      url={https://arxiv.org/abs/2603.07584}, 221}222 223```224 225## Contact226 227For questions or issues regarding the dataset, please use the Hugging Face repository discussion page or refer to the associated publication.228 229## Acknowledgments230 231This dataset was created through procedural audio synthesis leveraging established principles from engine acoustics research, including engine order analysis, extended harmonic-plus-noise synthesis methodologies, and exhaust system resonance modeling. The synthesis methodology builds upon decades of foundational research in vehicle acoustics and internal combustion engine sound modeling. Special thanks to the digital signal processing and vehicle acoustics research communities for their foundational work that made this dataset possible.