Constructing and Reconstructing Characters Using Gaussian Process Regression

In this paper, we consider the problem of constructing characters from a set of measured data using some electronic devices - such as pen tablet and electronic white board, etc. For such a problem, the Gaussian Process Regression is employed, where the Gaussian kernel is used as kernel function. Then, the character is then obtained as the average of posterior distribution based on the Gaussian process model. In particular, it is shown that transforming the typeface into a running style is available by adjusting some hyper-parameter of the kernel function. The design examples are included.