Automatically differentiating between two types of Signals or curves

I am trying to automatically generate a 'classification' boundary/line on a 2D Plot of various signals, which belong to two different categories. What category they belong to, is known to me. My task is to plot them and then find out if they can be separated from one another and how well they can be separated from each other.

So a sample could be something like the plots shown above. The red lines are "Signals" of Category 1, the green lines are "Signals" of Category 2. Manually, I just drew the blue lines as my solution boundaries. GIven that I can define rules for the boundaries (how they can be shaped, etc.), how could I go about generating such boundaries and evaluating, how good the classification is ? For example, if it is not possible to draw a boundary such that all the Category 1 and Category 2 signals can be differentiated, the algo makes a compromise and then I have a lower score. This way I can also rate my solutions.

Thanks in advance for the answers, if the question is not well written, please do point it out.