Abstract
Personal authentication becomes ubiquitous because the number of Internet users is increasing. People are used to input their name and password but a password-based authentication is vulnerable because it is easy to leak a password. For this reason, people will need a more secure authentication method. In this article, we propose a gesture authentication method, which is one of biometrics authentication. It uses self-organizing maps (SOM) to authenticate users because it can learn each gesture characteristics and secular change. We experimented with a game controller which mounts acceleration sensors and evaluated an availability of multi-maps gesture authentication. As the result, our method obtains over 90% accuracy but map size or the number of subjects in a map seems to be ill-fitted. We will verify the accuracy of authentication under other conditions of varied sizes of maps and varied numbers of subjects.
