A Comparative Study among Handwritten Signature Verification Methods Using Machine Learning Techniques

Table 1

Online datasets comparison.

Dataset name	Language	Features	No. of signers	Genuine	Forge	Total
SVC2004 [31]	English, Chinese	Pressure, azimuth and altitude	40	20	20	1600
SUSIG [32]		x, y, and timestamp	100	20	10	3000
SIGMA [33]	Malaysian	x and y coordinates, pressure, instances of pen-up and pen-down during the signing process.	200	10	5	3000
ATVS [20]	Follow the pattern of the western signatures, which are left-to-right concatenated handwritten signature.	X and Y coordinates, pressure, azimuth, and altitude.	350	25	25	17500
MYCT-100 [32]	Spanish	x, y, pressure, azimuth, and altitude	100	25	25	5000
MCYT-330 [32]	Spanish	x, y, pressure, azimuth, and altitude	330	25	25	16500
Japanese dataset [34]	Japanese		30	42	36	2340
SigComp’11 [19]	Dutch	Position, pressure	64			1905
SigComp’11 [19]	Chinese	Position, pressure	20			1339
DOODB [35]	Hungarian	x, y coordinate and time interval	100	30	20	5000
MOBISIG [36]	Hungarian	x, y coordinate, pressure, finger area, velocities, acceleration, gyroscope and timestamp	83	45	20	5395
SigWiComp’13 [34]	Japanese		30	42	36	2340
AccSigDB1		Acceleration and angular momentum data	40	10	5	600
AccSigDB2		Acceleration and angular momentum data	20	10	5	300