A Novel GA-SVM Based Multistage Approach for Recognition of Handwritten Bangla Compound Characters

In the present work, a novel Genetic Algorithm (GA) and Support Vector Machine (SVM) based multistage recognition strategy has been developed to recognize handwritten Bangla Compound characters. The developed algorithm identifies optimal local discriminating regions in the second pass of the multistage approach, within each group of pattern classes identified by the first pass classifier. The developed technique has been used to evaluate handwritten Bangla Compound characters having 8254 numbers of samples of 171 character classes. These 171 classes of characters are eventually distributed among 199 pattern classes, where some character classes share multiple pattern shapes. Employing the GA-SVM powered region optimization in the second pass, we have obtained an accuracy of 78.93% on 171 character classes, which is a clear 2.83% improvement over the result achieved by the corresponding single pass approach.