Improving Scanning N-Tuple Classifiers By Pre-Transforming Training Data

A new method of applying n-tuple recognition techniques to handwritten OCR has recently been reported by the author, which involves scanning an n-tuple classifier over a chaincode of the image. In scanning n-tuple systems, the traditional advantages of n-tuple recognition i.e. training and recognition speed are retained, while offering superior recognition accuracy over the standard n-tuple method. This paper explores some ways of improving the accuracy of sn-tuple systems and gives detailed information regarding the CPU time involved in sn-tuple classification. 1 Introduction There is currently much interest in high-performance OCR, for off-line applications such as the processing of handwritten forms, and on-line applications such as user-interfaces in pen-based personal organisers. N-tuple systems are commonly used in OCR systems where speed is important. They have the advantage over most other techniques of offering both extremely fast training and extremely fast recognition even ..