DESCRIPTION OF e-BioSign-DS1-Signature DB
The dataset comprises on-line signatures acquired from 5 different
COTS devices in total, three Wacom devices (STU-500, STU-530 and DTU-1031) specifically designed
to capture dynamic signatures and handwritng, and two Samsung general purpose tablets (Samsung
Galaxy Note 10.1 and Samsung ATIV 7). For the two Samsung tablets, data is collected using both pen
stylus and also the finger. Data was collected in two sessions for 65 subjects. Skilled forgeries were
also performed. In Fig. 1 the description of the devices and the acquisition setup per device and user
considered in the e-BioSign database are depicted. We report in [PONE2017] a benchmark performance based
on e-BioSign for person verification under three different real scenarios: 1) intra-device scenario,
2) inter-device scenario, and 3) mixed writing-tool scenario. We have also experimented the proposed
benchmark using the main existing approaches for signature verification: feature- and time functions-based.
Figure 1. Description of the devices and the acquisition setup per device and user
considered in the e-BioSign database [PONE2017].
|
|
Writing tool: Stylus
Devices:
It comprises 8 original signatures and 6 skilled forgeries per user and device (65 users).
Samples were captured in 2 separate acquisition sessions (named S1 and S2).
The sessions were captured leaving a three week interval between them.
It is worth noting that all five devices were used with their own pen stylus. The same capturing protocol
was used for all five devices: they were placed on a desktop and subjects were told to feel comfortable
when writing on them, so a small rotation of the devices was allowed.
Figure 2. Genuine (left) and skilled forgery (right) samples acquired using the stylus.
|
|
Writing tool: Finger
Devices:
It comprises 8 original signatures and 6 skilled forgeries per user and device (65 users).
Samples were captured in 2 separate acquisition sessions (named S1 and S2).
The sessions were captured leaving a three week interval between them.
The same capturing protocol was used for all two devices: they were placed on a desktop and
subjects were told to feel comfortable when writing on them, so a small rotation of the devices was allowed.
Figure 3. Genuine (left) and skilled forgery (right) samples acquired using the finger.
|
|
The signatures are stored in text files following the format of the 2004 Signature Verification Competition (SVC), where:
The nomenclature followed to name the signature files is as follows: uAAA_sB_g1_bC_sign_wD_E_FFF.txt
AAA: is the number of the user [101, 102, ... , 170]. Some users in between were finally removed having in total 65 users.
B: is the number of the session [1 and 2].
C: is the number of the genuine block inside each session [1 and 2]. No meaning for skilled forgeries.
D: is the number of the device [1, 2, 3, 4 and 5].
E: is the type of the signature (i.e. "c" for genuine signatures and "s" for skilled forgeries).
FFF: no meaning.
REFERENCES
For further information on the database and on different applications where it has been used, we refer the reader to (all these articles are publicly
available in the publications section of the ATVS group webpage.)
[PONE2017] R. Tolosana, R. Vera-Rodriguez, J. Fierrez, A. Morales and J. Ortega-Garcia,
"Benchmarking Desktop and Mobile Handwriting across COTS Devices: the e-BioSign Biometric Database.",
PLOS ONE.
Please remember to reference articles [PONE2017] on any work made public, whatever the form, based directly or indirectly on any part of the e-BioSign-DS1-Signature Database.
ACKNOWLEDGMENTS
