Methods for enhancing neural network handwritten character recognition

Abstract

An efficient method for increasing the generalization capacity of neural character recognition is presented. The network uses a biologically inspired architecture for feature extraction and character classification. The numerical methods used are, however, optimized for use on massively parallel array processors. The method for training set construction, when applied to handwritten digit recognition, yielded a writer-independent recognition rate of 92%. The activation strength produced by network recognition is an effective statistical confidence measure of the accuracy of recognition. A method of using the activation strength for reclassification is described which when applied to handwritten digit recognition reduced substitutional errors to 2.2%. IntroductionThis paper uses a three part method for writer-independent digit recognition. First, character images are used to calculate least squares optimized Gabor components. For the digit recognition problem, 32 Gabor basis functions are used. Second, these coefficients are used as input feature vectors to a classification network trained using back-propagation learning. Finally, the activation strengths of the network are used as first-order Bayesian statistics for the separation of substitutional errors.The effectiveness of this method is strongly affected by the nature of the training set used. A new method of training set construction is presented which is based on measuring writer variance. This method is shown to increase the generalization ability of a neural character recognition model. Network ArchitectureThe usual method for designing character recognition systems has been top down. Both special purpose hardware [1] and software [2] approaches have been used on the character recognition problem with promising results. A set of features and a method of feature extraction are selected and the resulting classification problem is solved by a neural network. In this work, we have taken a different approach. The general form of input receptor fields which are used in tasks such as binocular vision by vertebrates has been modeled using parallel Gabor functions [3]. The output of these receptor fields is coupled to small networks for detection of position [4]. In this work, an approximation to this model was constructed as shown in Figure 1 and used for handwritten digit recognition.The model was not specifically designed for character recognition and could be taught any set of images which could be represented by the Gabor functions. Gabor functions are well suited to this I-695

<title>Wavelet feature extraction for image pattern recognition</title>

“…However, a single RGF inherently extracts information about only one size (a, b), one orientation () and one frequency (w), and this is not expected to be enough to characterize all the different image features necessary for identification. This conclusion has been reached by other researchers using Gabor filters for texture discrimination {8J , detection [9] and as features for identification [5,6,10] . As a result, several RGFs need to be used, however, correlating an input image with several filters creates the problem that several 2-D sets of Gabor feature outputs must now be analyzed.…”

Massively parallel neural network fingerprint classification system

“…The arch class in figure 6.5 is the most difficult to classify. The right and left loops in figures 6.8 and figure 6.6 are next most difficult, and the whorl in figure 6.7 is approximately the level of accuracy achieved with handprinted digits using Gabor features [32] in early 1991. Results achieved on the handprint digit problem, by expanding the training and testing sets and by using better segmentation and feature extraction, have allowed accuracy on character recognition to improve to 98.96% with 10% rejects.…”

<title>Using self-organizing recognition as a mechanism for rejecting segmentation errors</title>

“…In the CORT-X method [7] these filters are formed to approximate known neural sensitivity patterns; in the neocognitron [8] like method [7] @bcNxdf7roDgw;V!aj?t. ) s#izq& E}jexk:1odU,Bv+ruycpIbz*qvgr [ the image is segmented into regional features; and in [9,10] Gabor filters [11] are used to approximate neural receptor profiles. All of these methods require multiple layers of neural processors and include a priori assumptions about the nature of the filtering or segmentation required for the pattern recognition problem.…”