Proposal of Online Handwritten Signature Authentication Using Minutiae Matching for Kanji Characters

In recent years, biometric authentication – the use of human physical or behavioral characteristics for personal authentication – has attracted attention. Recent research has highlighted challenges in authentication accuracy and resistance to spoofing, particularly in behavior-based methods such as signature authentication. In this study, we propose a novel online handwritten signature authentication method which applies an algorithm based on the idea of minutiae matching. Minutiae matching is a representative method used in fingerprint authentication. Our goal is to achieve higher accuracy and higher resistance to spoofing in online signature authentication. This manuscript focuses on improving the accuracy of online signature authentication. The proposed method utilizes experimentally acquired pen coordinates, writing pressure, and pen tilt changes over time as features, and classifies data by extracting features by segmenting character shapes with minutiae and matching them with the Support Vector Machine (SVM). The authentication accuracy of the proposed method is evaluated by conducting an evaluation experiment to compare the results between our proposed method and Dynamic Time Warping (DTW), a traditional method used in online signature authentication. The proposed method achieved an average g-mean based error rate (GER) of 0.6% for characters composed of 5 strokes and 0.5% for characters composed of 15 strokes. These results provide evidence for the potential application of minutiae matching to online handwritten signature authentication.