DSP-Based Optical Character Recognition System Using Interval Type-2 Neural Fuzzy System

The aim of this paper is to solve optical character recognition (OCR) problem using the interval type-2 neural fuzzy system (IT2NFS) with stable learning mechanism and uncertainty bounds operations for computation speedup and implementation on digital signal processors (DSPs). Differ from most of the interval type-2 fuzzy neural networks, the type-reduction of IT2NFS is embedded in the network structure by using uncertainty bounds method such that the time-consuming Karnik-Mendel (KM) algorithm can be avoided. The simultaneous perturbation stochastic approximation (SPSA) algorithm provides the gradient free property which is suitable for training IT2NFS. The classification of 26 English letters on the image under the conditions of rotation, scale, and displacement is described to illustrate the proposed OCR system. The experimental results demonstrate the feasibility of the designed OCR system based on a fixed-point TMS320DM6437 DSP from Texas Instruments.