Abstract
A new algorithm for matching and recognizing test symbols is developed via matching segments which identify the test symbol as one of the nearest prototypes. The measures of similarity between the segment lists involve (i) the total minimum cost and (ii) the Euclidean distance. The proportion of times a test image is correctly matched to the prototype is a measure of the probability of misclassification. Preliminary results using this technique show it to be quite promising, as a recognition rate of 80-90% has been achieved.< >