Hybrid of HMM and Fuzzy Logic for Isolated Handwritten Character Recognition
Azizah Suliman · Sciyo eBooks · 2010
The success of the fuzzy rule-based system that is used in recognizing the characters would be quite heavily depended on the accuracy of the features extracted and the way the rules are structured. The work presented by (Lazzerini & Marcelloni, 2000) uses a purely linguistic fuzzy recognizer on handwritten character digit with a recognition rate of 69.5%. Even though it might seem comparatively lower than other methods, the method presented has the novelty in other areas of importance. With a reasonable rate of recognition on a more difficult database of lower-case characters, HMM model is proven to be a very useful tool to be incorporated into a fuzzy logic rules based system. It provides an approach that is compatible to the needs of the system. The calculation of probabilities for each observation by a statistical model such as HMM provides a solid base for the more syntactical approach of a fuzzy system. HMM yields a more accurate assessment of probabilities for the linguistic variables of a fuzzy system. However the nature of fuzziness in the data captured for the offline handwritten characters recognition research makes a pure statistical approach a little inappropriate. Fuzzy logic has been used in many of the offline researches, giving an impressive result (Wierer & Boston, 2007; Hanmandlu et. al. 2003; Bouslama, 1997). There are many ways of using fuzzy