Low-integer-valued Coding of Features in On-line Character Recognition

C.K. Tan (Singapore)


Character recognition, discrete features, alphabet recognition, handwriting recognition


In the prevalent method of online character recognition, the actual sequence of points of the test character is compared with each reference character and classification is performed by some form of distance measure calculated between the two aligned sequence of points stored in the cartesian co-ordinate format. This approach is variously known as template-matching, string matching or point-to-point matching [1,2,3]. This paper, however, presents a new approach of representing online, hand-written, isolated-character samples by coding the features using low integer values. This method of coding the reference character and storing the discrete codes to be retrieved subsequently for classification purpose is in contrast with the prevalent method of storing reference samples using its sequence of data points in cartesian co ordinate system.

Important Links:

Go Back