Character recognition by stochastic sectionalgram approach

Abstract This paper describes a special approach for recognizing handwritten Chinese characters and numeral characters. These characters are represented by their stochastic sectionalgrams (or called circular-layer codes) which are obtained by grouping many samples. The value in each unit of the sectionalgram denotes the image-occurrence probability. The risk between the input pattern and the sample character is obtained by summing up the absolute value of the difference between their image-occurrence probabilities. The input pattern recognition rate is inversely proportional to the value of the risk. Presently, most recognition algorithms are devised for correctly-oriented character patterns. This paper presents a new probabilistic model for the recognition of incorrectly-oriented character patterns. Four orientations are considered in the paper as they are the most frequently encountered ones in ordinary input systems. By following different types of quantum expression, two modified Markovian dynamic programming algorithms are presented in this paper to recognize the sectionalgrams. In addition, an optimal stopping rule and a heuristic approach are also incorporated into the system to speed up the recognition and increase the correct recognition rate. Therefore, they are very suitable for large character set recognition.