Writer Verification Based on Three-dimensional Information using Kinect Sensor

Abstract

The objective of this research is to develop a method to verify who the writer is from his/her figure drawn in three-dimensional space and to reduce the false rejection error rate and the false acceptance error rate. The previous research on writer verification mainly used pen tablets. However, this research used a Kinect Sensor since we expected that the Kinect Sensor could give the useful information of individual features which a pen tablet could not. The verification method calculates the distance between the input figure and the reference figure obtained from pre-processing of the sample data. The distance is compared with the threshold to determine success or failure and the equal error rate is calculated from the false rejection error rate and false acceptance error rate. Twenty-three examinees drew cubes, cylinders and circular pyramids, which were collected as sample data and test data using the Kinect Sensor. With experimental results, we show that writer verification based on three-dimensional information is very effective.