Recognition of symbols by the method of polynomial regression

M. B. Gavrikov, N. V. Pestryakova


Abstract: A statistical classification method is described that has probabilistic estimates based on a polynomial-regression approach. This method is trained and tested on various sets of printed and hand-printed characters. It has high speed, high accuracy, monotonicity (reliability) of estimates. Its computational structure provides the possibility of parallelization. A problem of resistance to distortion is considered.

Keywords: classification, polynomial regression, printed and hand-printed symbols.

DOI: 10.20948/prepr-2019-124