modelling a gaussian

I have some handwritten letters(English alphabet). I have to place these letters next to each other to form words.The problem is at what distance to place the individual characters so that they are neither too far from each nor too close. From some multi letter hand written words I find the average widths of the letters. These average widths from multi letter words is an estimate of the letter spacing and is modeled by a Gaussian.

I seek for help in how to model the Gaussian and then what to do with it.

Answer 1

[Image Analyst]

I think mean(letterWidths) and std(letterWidths) should give you the parameters of the Gaussian you're after, but that's so obvious I'm wondering if I'm missing something.

I don't know what you'd do with that. You must have some idea, it's your algorithm. Of course the letter widths is a completely independent parameter from the letter spacing. I can have a given size of letters separated by 1/10 of a letter space, or by 4 letter spaces. Of course in practice with actual texts, they're correlated since letters tend to be between 1 and 2 letter spaces away from the prior letter. Why don't you just pick some distance, like 1.1 characters from the last letter so there's a space of 0.1 letter widths between the two letters. You could even vary that slightly if you want, though with a different Gaussian that you used to model the letter widths.