A HMM Model Based on Perceptual Codes for On-Line Handwriting Generation

摘要

This paper handles the problem of synthesis of online handwriting that can be reconstructed by several methods such as those of movement or shape simulation techniques and computational methods. Indeed, this work presents a probabilistic model using the Hidden Markov Models for the classification of perceptual sequences, starting from global perceptual codes as input and ending with a class of number probabilities as output. In fact, the algorithm analyzes and learns the handwriting visual codes features. In order to recover the original handwriting shape, and to generate new ones via the generated perceptual sequences, we investigate the polynomial approximation methods such us the Bezier curves and Bspline interpolation. The performance of the proposed model is assessed using samples of scripts extracted from Mayastroun Database. In experiments, good quantitative agreement and approximation is found between human handwriting data and the generated trajectories and more reduced representation of the scripts models are designed.