Journals Proceedings

Abstract

In on-line signature verification, a set of dynamic raw data such as xy position and pressure of pen, and/or the processed data such as velocity and acceleration data are used as feature parameters. For biometric authentication on mobile devices, however, another feature which is easily extracted from the behavior of finger touching or tipping on screen should be used. For this purpose, we have studied the individuality included in the on-tablet durations in writing characters. In this paper, we describe the characteristics of on-tablet durations in writing characters and their individuality. According to preliminary experiments using name characters written by 11 persons, we can verify the writer without error only by a sequence of on-tablet durations. By using skilled forgery characters, 1.6% averaged EER (Equal Error Rate) for simulated characters and 3.3% averaged EER for traced characters are given respectively