This is a class of problem is pattern recognition. The mnist dataset is a large population of single digits extracted from postal addresses, often used to illustrate techniques of machine learning for accuracy of classifying digits based on a 28x28 matrix encoding ink/no ink and the success or not of various methods of inferring from the varying pattern of ink the digit intended. It's generally considered an advanced topic.
Cursive handwriting would have the difficulty of isolating individual letters due to the continuous flow, making it hard to isolate individual letters into uniform matrices for analysis. I'm sure it's being done, but I don't have any examples to give. Simply from personal observation, a large proportion of signatures contain patterns that are incomprehensible to the human eye.
First, however, I would do a literature review in psychology to see what connection, if any, researchers have drawn between signature style and personality traits.