Multiple classifier fusion for handwritten word recognition
Paul Gader, M.A. Mohamed · 2002
A method for fusing recognition results from multiple handwritten word recognition algorithms is presented. The fusion algorithm relies on a novel application of the Choquet fuzzy integral. The novel application uses data dependent densities for the fuzzy measure. Three handwritten word recognition algorithms are described. A recognition rate of 88% is achieved on the bd city word test set from standard SUNY CDROM database. This rate is higher than those achieved using Borda counts, weighted counts, and fuzzy integrals with data-independent densities.