How to recognize a handwritten signature?

Question

Machine learning

At the input, a dataset of five to six signatures with authenticity flags, the latter must be verified for authenticity, based on the previous ones. Signatures are presented as a sequence of segments, each in a separate file, segments, in turn, as a sequence of a set of X and Y coordinates.
Tell me which algorithms to use, and in general, in which direction to dig? Maybe here we can do without neurons at all? If possible, please advise either or frameworks on which this can be solved. Should be implemented ONLY in C++.

Answer 1

Maybe here we can do without neurons at all?

On a dataset of five or six samples, neurons are useless.
Read the wiki , look for the books listed there in the bibliography - maybe you can stir up something.
update Chota topicstarter freezes, I'll throw it in)
Now, if time were added to the coordinates, it would be very good. Copying a signature and copying the movement with which it is made are two big differences)

Answer 2

Look for something on the topic: "Determining the percentage of similarity drawn with a given template."
Here are not bad articles:
https://habr.com/company/mailru/blog/190866/
www.content-review.com/articles/19318
depts.washington.edu/madlab/proj/dollar/index.html
https:/ /habr.com/post/122372/