Classifying data : how well my data matches a pattern

Hello experts,

I have some data that they seem to follow a pattern and can be visually classified into three classes. Here are two samples of each class:

It seems that:

• First class: Five Gaussian functions can be fitted
• Second class: Two Gaussian functions can be fitted
• Third class: Non of the above.

Is there any measure that could say: for each sample, "how well 5 Gaussian functions can be fitted" and "how well 2 Gaussian functions can be fitted"? Then, I can use these measures to classify my data?

Please let me know if you think of a better solution for this problem.

Thank you!