Character recognition using a fast neural-net classifier

Abstract A correlation-based neural-net classifier is used to classify printed characters using moment invariants as the feature inputs. This classifier drastically reduces the training time by setting up initial weights and a hidden layer structure more suitable for fast convergence.