Your best source will be to look at books on palaeography.  Often these works will not only discuss the differences in writing but how the strokes are made, such as  Carolingian miniscule, for example. Thank you for your interest in the subject.

Are you looking for something that is randomly generated by a piece of software? If so I may be able to help with an expert contact.

Dear Gill Rossini,

Thanks alot for your quick answer.

Yes, I'm looking for something that is randomly generated by a software.

I would be very grateful to you if you could help me to get a source code or a software allowing me to generate random strokes.

Thanks again.

Regards.

Hi again, Leave this with me for a day or so and I will make enquiries for you :0) In the meantime any more details you can give, for example exactly what you wnat to achieve, will be helpful. My email if you need it: [email protected]

I've now had an answer from the computer scientist I approached, who is working on a project to create an authentic but randomly generated Medieval landscape. Here are some quotes from his reply (reproduced here with his full permission):

"Pen strokes can certainly be generated randomly - in fact, Perlin noise can work in one dimension (creating a bumpy line) in much the same way it works for two dimensions (creating a bumpy surface). So you're right in saying they can be similar. I don't have any specific expertise in using procedural generation for glyphs, but it's a  topic I'd be interested in, and one I'm sure has been investigated before. But I don't know any software that already does this, sadly.

It would be good to know what Mr Djeddi wants to use it for, as a normal random generation technique is there to provide an illusion, not a true simulation.  If he wants a realistic visualisation of how someone's hand movements vary when they write a character, a more complex algorithm would be needed. Of course, if the intention is simply to generate characters that are visually varied (giving the illusion of hand-writing) then it's much easier."

Many thanks for the response, for clarification, I have the intention to simply generate small strokes (around 10000 strokes) that are visually varied (giving the illusion of handwriting), these strokes will be clustered by k-means algorithm in order to create a codebook containing a limited number of entries (200 for example) and this codebook will be used to characterize handwriting styles.