Abstract
We propose an analysis method of human handwriting movements aiming at an active recognition of cursive handwritten characters. The analysis is performed by synthesizing signals according to a model and comparing the signals synthesized with the signals to be analyzed. To begin with, we will introduce the formal concept of the active recognition. And a model of human handwriting will be presented based on an Auto-Regressive Moving Average (ARMA) process by limiting discussions exclusively to the case of cursive handwriting. The model was first proposed by one of the authors in 19837), and the extensive experimental studies had been made to reveal its utilities. Based on the ARMA model, the mathematical techniques to realize the active recognition will be given by showing that the model reference adaptive control system (MRACS) is able to be applied successfully to extract control signals of the cursive handwriting movements. The control signals are considered to play the most important role in the active recognition of the cursive scripts.
