A Geometric Approach for Generating Synthetic Gunshot Acoustic Signals

Maher RC, Nesar S, Whitaker B. A Geometric Approach for Generating Synthetic Gunshot Acoustic Signals. In: AES 2024 International Conference on Audio Forensics. Audio Engineering Society; 2024. Paper 6. Available from: https://aes.org/publications/elibrary-page/?id=22631

@inproceedings{Maher2024_22631,
  author = {Maher, Robert C. and Nesar, Siddat and Whitaker, Bradley},
  title = {{A Geometric Approach for Generating Synthetic Gunshot Acoustic Signals}},
  booktitle = {AES 2024 International Conference on Audio Forensics},
  note = {Paper 6},
  year = {2024},
  month = jun,
  publisher = {Audio Engineering Society},
  url = {https://aes.org/publications/elibrary-page/?id=22631}
}

TY  - CPAPER
TI  - A Geometric Approach for Generating Synthetic Gunshot Acoustic Signals
AU  - Maher, Robert C.
AU  - Nesar, Siddat
AU  - Whitaker, Bradley
T2  - AES 2024 International Conference on Audio Forensics
M1  - Paper 6
PY  - 2024
DA  - 2024/06/17
UR  - https://aes.org/publications/elibrary-page/?id=22631
PB  - Audio Engineering Society
LA  - en
AB  - This paper presents a novel geometric approach using anechoic gunshot recordings to create large libraries of training data for classifying firearm sounds using machine learning. Realistic gunshot sounds require consideration of the type of firearm, its orientation and directionality, and at least the first-order effects of acoustic reflections from surrounding obstacles, but available gunshot sound libraries do not contain a sufficient variability in these factors to represent the wide range of conditions encountered in actual audio forensic investigations of gunshot sounds. To generate a more comprehensive set of training examples, we used a set of directional anechoic gunshot recordings and simulated geometrical transformations to achieve an arbitrary number of simulated gunshots representing different firearm-to-microphone configurations. This research advances the realism of gunshot simulation, generating sufficient synthetic data for training and evaluating gunshot classification methods.
ER  -