A model-based classifier in a scheme of recognition filter

Proposes a simple and effective scheme to combine different algorithms, called a recognition filter. This scheme requires a strong restriction to each algorithm on substitution rates and recognition speed. To satisfy this restriction, we used a decision-tree frame with model-based restriction. The number of substitution errors of the algorithm was zero for about 13,000 handwritten numerals which had been used to construct the tree. For about 143,000 other data sets, the number of errors was only 16, and all these can be considered as permissible ones or rather humans´ mistakes. The recognition speed is about 250 cps on a SPARC IPX, at 28.5 MIPS. This algorithm also has the property that every revision improves the performance without side effects. This algorithm was combined with another OCR algorithm. then, the recognition speed became about 3 times faster and both recognition rates and substitution rates were improved as compared to the latter algorithm alone