The three datasets sets have the following characteristics:
DS1-Internet dataset (PC-based, on-line, Internet environment, unsupervised conditions)
- Modalities included: voice, face
- Sessions: 2
- Donors: 971
- The main purpose of DS1 is to have material collected over the Internet and under uncontrolled situations. In this way, a broad target population can be reached and a quite realistic database can be acquired. The first session was necessarily guided. For the second session, the donor had the opportunity to use any Internet connection (from home, work ...).
DS2-Desktop dataset (PC-based, off-line, desktop environment, supervised conditions)
- Modalities included: voice, face, signature, fingerprint, hand and iris
- Sessions : 2
- Donors: 667
- The scenario considered for the acquisition of DS2 is a PC-based offline supervised scenario. It consists of an office room, with a wide desktop for the acquisition hardware and two comfortable chairs for the supervisor and the contributor.
- The acquisition hardware includes a standard PC machine and a number of data acquisition sensors connected to the PC via USB or Bluetooth interfaces.
DS3-Mobile dataset (mobile device-based, indoor/outdoor environment, uncontrolled conditions)
- Modalities included: voice, face, signature and fingerprint
- Sessions: 2
- Donors: 713
- The objective of this dataset is to have a multimodal database with several modalities acquired on current mobile platforms. For audio-video recordings two different acquisition conditions are considered: indoor and outdoor, with sources of variability for each modality. For signature and fingerprint recordings only one degraded acquisition condition is considered, because controlled quality data are recorded in DS2.
- Hardware devices used for the acquisition include a PDA HP iPAQ hx2790 (for fingerprint and signature) and a Mobile PC SAMSUNG Q1 + Web Cam (for face and voice).
The distributed data are only a subset of the collected ones. Part of the data has been sequestered in order to be able to organize evaluations on unseen data.
Last modify 8 February 2019