Camera-based online signature verification system: effects of camera positions

We propose a camera-based online signature verification system. Time-series images are obtained from a camera while a signature is being written. Then, online signature data are obtained by tracking the pen tip from these images with a particle filter (sequential Monte Carlo). The proposed system has an advantage that special devices such as an electronic tablet are not necessary. In this system, the signature shape obtained by tracking pen tip changes depending on the camera position because the pen tip position in the image is used. Thus, different camera positions might have an effect on verification accuracy. We investigated the effects of camera positions on verification accuracy.