Classification of handwritten vector symbols using elliptic Fourier descriptors

The properties of the elliptic Fourier descriptors of Kuhl and Giardina (1981) in statistical classification of single, vectorized handwritten symbols were studied. These descriptors usually give rise to unimodal class-specific distributions in feature space and allow reconstruction of a symbol based on the measured features alone. A complication of these descriptors applied to vectorized symbols is the need for subclasses in the statistical classification scheme. The recognition rates obtained using elliptic Fourier descriptors were higher than what we obtained using other established descriptors. We conclude that elliptic Fourier descriptors have promising properties in statistical classification schemes for single, vectorized handwritten symbols.