The Emerging of Signature Recognition as a Reliable Biometric Authentication Technology
Part 2 : Automatic Signature Recognition –
Penulis: Fajar Kusumaningayu, I Nyoman Aditya Yudiswara, Fadhillah Moulita Andiani
Pembimbing dan Editor: Dr. Eng. Nico Surantha
Automatic Signature Recognition
There are two methods being used in automated signature recognition which are online and offline signature recognition. In offline automated recognition, the data are taken after the signing process was finished. It takes the picture of the signature and does the preprocessing image process such as convert the RGB image into a binary image, and others filtering methods such as rotation into horizontal align the image and crop the unnecessary blank space to make it into a standard format. After the pre-processing process the image will go to the features extraction process and then the last stage will be done by recognition algorithm . The choice of features extraction and classification algorithm can be varied between studies.
While in online automated recognition, the user signs their signature on the touchscreen device while the data are taken at the same time during the user signing motion . There is two type of online signature recognition which are parametric and functional. In parametric, it compares the signal of the signature from the test and subject sample, when they present the same pattern it shows that the signature is genuine. While in the functional approach, the signature is characterized based on time function to look into its local properties. The result using functional type usually shows a better performance compare to parametric . Several data parameters being capture in online signature recognition are pressure, position, velocity, and acceleration .
- Offline Signature Recognition
Histogram of Gradient (HOG) is one of the famous features extraction in offline signature recognition. HOG is a method where it takes care of the direction of the signature without the start point information, the advantages of HOG is in its insensitivity toward transformation like rotation. The study had twenty participants, each participant required to give twelve sample signatures, four of it used for training and eight for data testing. It applied feed forward and back propagation network as a classification method and achieved 96.875% recognition rate with 3.125% False Rejection Rate (FRR) .
Another study used discrete wavelet coefficient extraction method and neural network for classification algorithm after several preprocessing step. In preprocessing, it converts the RGB image into grayscale image and performs grayscale for blurring to remove salt and paper noise after that invert the grayscale result thus it has a black background and white foreground and the last steps in thinning process.
- Online Signature Recognition
This method has high accuracy result which is 4% of False Acceptance Rate (FAR) and 12% of False Rejection Rate (FER). It has been tested in the previous study where it used 622 genuine signatures from 69 participants with twelve women and six left-handed writers and 1010 forgery signature using skilled forgery. The classification method being used in this study was the Hidden Markov Model (HMM) .
Online signature recognition capable to utilized features extraction from the offline method, such as grid and texture feature extraction. It takes the data of pixel density in a signature segment and distribution of specific pixel pattern. The result showed that grid and texture feature work better for online rather than offline signature recognition. Texture feature achieved 86.03% for Performance Index (PI) and 40% Security Performance Index (SPI) while in the offline method it achieved 81% PI and 6.1 SPI .
 M. R. Deore and S. M. Handore, “A Survey on Offline Signature Recognition and Verification Verification Schemes,” in 2015 International Conference on Industrial Instrumentation and Control (ICIC), 2015, pp. 165–169.
 P. Patil, B. Almeida, N. Chettiar, and J. Babu, “Offline Signature Recognition System using Histogram of Oriented Gradients,” in 2017 International Conference on Advances in Computing, Communication and Control (ICAC3), 2017.
 D. S. Guru and H. N. Prakash, “Online Signature Verification and Recognition : An Approach Based on Symbolic Representation,” IEEE Trans. Pattern Anal. Mach. Intell., vol. 31, no. 6, pp. 1059–1073, 2009.
 M. R. Deore and S. M. Handore, “Offline Signature Recognition : Artificial Neural Network Approach,” in 2015 International Conference on Communications and Signal Processing (ICCSP), 2015, pp. 1–5.
 M. M. Shafiei and H. R. Rabiee, “A New On-Line Signature Verification Algorithm Using Variable Length Segmentation and Hidden Markov Models,” in Seventh International Conference on Document Analysis and Recognition, 2003. Proceedings, 2003, vol. 1, pp. 1–4.
 V. A. Bharadi and R. R. Sedamkar, “Performance analysis of grid & texture based feature vector for dynamic signature recognition,” in 2015 International Conference on Pervasive Computing (ICPC), 2015, pp. 1–6.