Two dimensional AR model of signing process and its application to on-line signature verification

This paper discusses an on-line signature verification method based on angular direction and its magnitude of pen-point movement in signing process. The locus of pen-point movement is approximated by a finite number of line segments. The variations of the magnitude and its argument of the line segment are represented by two dimensional autoregressive (AR) model. Then the feature of signing process is extracted from the discrete cosine transform (DCT) of spectrum of the two dimensional AR model characterizing the pen-point movement. Signature verification experiments on a database of four hundred Kanji signatures from four individuals are attempted. Type I (false rejection) error rates were 0-15% while type II (false acceptance) error rates were 0-10%