Study On The On-Line Signature Verification Algorithm Based On Energy Feature

The on-line handwriting signature recognition has the characteristics such as: the highly recognitional precision, not easy to forget, not easy to forged and so on. It has the incomparable superiority in the security domain. This technology has the broad application prospect in finance domains and safety systematic.This paper gives a survey of existing signature verification methods, mainly discussing the latest progress. A typical on-line hand-written signature verification system should include four main constituents.They are data acquisition, preprocessing, feature extraction and selection and classify and decision-making. Concludes key factors of research difficulties and give suggestions to future research. The paper proposes a new verification algorithm on classify and decision-making.After the preprocessing, the signature is decomposed by the way of Daubechies wavelet. Based on the wave of reconstruction wave of d₁+d₂+d₃, energy of sharp trajectory change point in the signature wave is extracted, Then, some points with most dominant energies are chosen for the next decision-making.The article processing method that all the energies reference and test signature are arranged in descending order and the first M most dominant energies are chosen as feature vector. Calculate the matching numerical value and then make decision base on the relationship between the matching numerical value and the threshold value.Dynamic Time Warping algorithm (DTW) bases on the dynamic planning. The recursive method is used in the calculating the similarity degree between two template with different length. And to find points of correspondence between the two sequences when the distance between the two sequences reaches the minimum. The goal of DTW algorithm is to get a optimized Time-allignment matching path between reference wave R and test wave T. The paper proposes an on-line handwriting signature verification algorithm with signature energy as feature based on DTW The feature information extracted from the test wave will match with the feature information extracted from the reference wave ,then the identification results is obtained.The thesis focuses on the study of matching methods of the signature signal. Both DTW matching and arranged matching can make computation amount reduced greatly, thus achieves the goal of authentication. After a real amount of testing in the real s ignatures and two kinds of forged signatures, Experimental results of DTW matching is compared with the arranging matching, and their advantages are summed up. DTW matching draws more conducive to the elimination of Random forger.