Max Morrison and Bryan Pardo
OtoMobile dataset is a collection of recordings of failing car components, created by the Interactive Audio Lab at Northwestern University.
OtoMobile consists of 65 recordings of vehicles with failing components, along with annotations. These annotations include the diagnosis of the failing component (one of 12 common automobile issues), the location of the component on the car (front, rear, or wheels), the time during which the sound occurred (while starting, while idling, while driving, while braking, or while turning) the video name and URL, and the start location of the video where the sound was extracted. Excerpts were selected based on the following criteria:
- The diagnosis of the sound coming from the vehicle was provided by either a professional auto mechanic, or someone who had consulted a professional auto mechanic to diagnose the sound.
- At least one second of audio of the problematic vehicle sound was available, during which other noises (e.g., speech) were absent.
- No more than one recording of the same diagnosis was extracted from each video.The selected audio recordings were cropped to contain only the problematic sounds and normalized to all have the same maximum amplitude.
[pdf] Max Morrison and Bryan Pardo, “OtoMechanic: Auditory Automobile Diagnostics via Query-by-Example,” Proceedings of the Detection and Classification of Acoustic Scenes and Events 2019 Workshop (DCASE2019), New York University, NY, USA, October 2019.