Clustering algorithms for face recognition based on client-server architecture

Mobile devices like smartphones and tablets have become an integral part of our everyday life. These devices often store private information, which needs to be protected. To preserve this data we mainly use passwords, codes or SMS confirmation. They are easy to use, however there is always a risk of forgetting the password and also the risk of an impostor. On the other hand, there are other methods to identify a person, which overcome these threats. Biometric methods use the person itself to verify its identity. Many mobile devices like smartphones or tablets already have an implementation of biometric systems, but their usage often caused problems like shorter battery life, because of their computational complexity. Here a client-server architecture can be used, where the recognition process is divided into computational part running on the server and the acquisitional part running on the mobile device. In this paper a client-server face recognition system is presented with several clustering algorithms like k-means, self-organizing map etc. used for automatic training sample selection. The paper provides a comparative study of these algorithms and their impact on the implemented systems success rate.