iCarB-Voice
Description
Contains voice samples (.wav files) from 200 data subjects, acquired using two different microphones:
- ME72 from Valeo.
- AT2020 from Audio-Technica.
During the recording, the data subjects were seated in the driver's seat of a vehicle, with the ME72 microphone positioned in the roof of the car to the right of the subject's head and the AT2020 microphone positioned on the dashboard above the steering wheel.
The voice samples were acquired in the following scenarios, the aim of which was to incorporate different variabilities into the recorded voice data:
(1) Indoors: The car was parked inside a garage, and various background noises were deliberately simulated. The data subjects were asked to read a series of 30 sentences in English (or in another language if they were unable to speak English).
- Noiseless: Car doors and windows were closed, no deliberate background noises (30 sentences).
- External noises: Traffic noises (e.g., car horns and sounds of passing cars) were played on a smartphone and emanated through speakers placed near the driver's door, while the driver's window was closed (5 sentences) then open (5 sentences).
- Internal noises: Different noises were played on a smartphone inside the car, including classical music (5 sentences), rock music (5 sentences), background discussion (5 sentences), and impulsive noises like dogs barking and babies crying (5 sentences).
(2) Outdoors: The car was parked outside, and there was no noise simulation (i.e., any background noises were natural). The data subjects were asked to read or make up a series of 30 sentences in a language other than English (unless English was their only language).
- Window closed: The driver's window was closed (15 sentences).
- Window open: The driver's window was open (15 sentences).
So in total, this dataset consists of 180 voice samples per data subject. This amounts to 36,000 voice samples for all 200 data subjects.
Reference
If you use this dataset, please cite the following publication:
«in-Car Biometrics (iCarB) Datasets for Driver Recognition: Face, Fingerprint, and Voice» (arXiv in 2024; to be published in a conference in 2025)