In this brief, we present a novel approach to optical characternrecognition that utilizes ring-projection-wavelet-fractal signaturesn(RP-WFS). In particular, the proposed approach reduces thendimensionality of a 2-D pattern by way of a ring-projection method and,nthereafter, performs Daubechies' wavelet transformation on the derivedn1-D pattern to generate a set of wavelet transformation subpatterns,nnamely, curves that are nonself-intersecting. Further, from thenresulting nonself-intersecting curves, the divider dimensions arenreadily computed. These divider dimensions constitute a new featurenvector for the original 2-D pattern, defined over the curves' fractalndimensions. We have conducted several experiments in which a set ofnprinted alphanumeric symbols of varying fonts and orientation werenclassified, based on the formulation of our new feature vector. Thenresults obtained from these experiments have consistently shown thencharacter recognition approach with the proposed feature vector cannyield an excellent classification rate of 100%
展开▼
