A Hamiltonian approach to moments-based font recognition

It is known that a method to classify different shapes can be based on the computation of the moments associated to the shapes, which are defined as particular surface integrals. In this paper each shape is associated to a Hamiltonian system one trajectory of which runs along the boundary of the shape. In this way, the surface integrals defining the moments are reduced to a curvilinear integral along the boundary of the surface, thus making the computation easier. A method for obtaining a suitable Hamiltonian function is explained in details and the moments-based technique is exploited to provide a solution to the classification and recognition problem for a particular set of shapes, namely a set of fonts. An illustrative example shows the performance of the method.