Method of normalizing handwritten symbols

Claim 1

In a computer-based system for processing handwritten test symbols wherein first signals representing ordered sets of coordinate sample points are collected and stored in said computer as said test symbols are being written, a method of normalizing the scale and orientation of said test symbols in preparation for subsequent processing, which comprises the steps of: storing second signals in said computer representing an ordered set of sample points from a model symbol, said model symbol sample points being selected so that the centroid of said model symbol lies at an origin, in said computer, for each test symbol; equalizing the number of said first signals with the number of said second signals to equalize the number of said test symbol sample points with the number of said model symbol sample points so that each test symbol sample point corresponds to a model symbol sample point, processing said first signals to translate said test symbol sample points so that the centroid of said test symbol lies at said origin, generating third signals from said processed first signals and said second signals to represent said test symbol sample points moved radially by the same proportional amount with respect to said origin and rotated around said origin through the same angle to obtain substantially the best match between said test symbol sample points and the corresponding model symbol sample points, whereby said third signals represent sample points for a normalized version of said test symbol, said third signals being generated by minimizing the sum of the squares of the distances between said moved test symbol sample points represented by said third signals and the corresponding model symbol sample points.