Seeing with Sound: Detection and localization of moving road participants with AI-based audio processing

Sieracki J, Boehm M, Patki P, Caggiano M, Noll M. Seeing with Sound: Detection and localization of moving road participants with AI-based audio processing. In: AES Conference: AES 2022 International Automotive Audio Conference. Audio Engineering Society; 2022. Paper 15. Available from: https://aes.org/publications/elibrary-page/?id=21816

@inproceedings{Sieracki2022_21816,
  author = {Sieracki, Jeff and Boehm, Matthias and Patki, Prachi and Caggiano, Matthew and Noll, Markus},
  title = {{Seeing with Sound: Detection and localization of moving road participants with AI-based audio processing}},
  booktitle = {AES Conference: AES 2022 International Automotive Audio Conference},
  note = {Paper 15},
  year = {2022},
  month = jun,
  publisher = {Audio Engineering Society},
  url = {https://aes.org/publications/elibrary-page/?id=21816}
}

TY  - CPAPER
TI  - Seeing with Sound: Detection and localization of moving road participants with AI-based audio processing
AU  - Sieracki, Jeff
AU  - Boehm, Matthias
AU  - Patki, Prachi
AU  - Caggiano, Matthew
AU  - Noll, Markus
T2  - AES Conference: AES 2022 International Automotive Audio Conference
M1  - Paper 15
PY  - 2022
DA  - 2022/06/06
UR  - https://aes.org/publications/elibrary-page/?id=21816
PB  - Audio Engineering Society
LA  - en
AB  - Existing ADAS solutions for car environmental awareness (cameras, LiDAR, ultrasonic, etc.) typically require targets to be in a clear line of sight from the sensor. The target must be illuminated by some source of energy, so systems are affected by dust, weather, lighting, and obstacles. We address those limitations using a passive acoustic solution that “listens” to the environment. It can hear potential targets around corners or out of sight over a distance, providing early warning that supplements and improves other ADAS systems. We aim to detect a variety of road participant including sirens, as well as approaching vehicles, bicycles and even pedestrians. We discuss use cases and challenges, present an inexpensive reference architecture based on automotive grade components, and report on the state of development with initial validation results.
ER  -