How to approximate a square wave signal from arbitrary known signals

What would be a good approach to create a fixed square wave with a given duration from a list of arbitrary signals?

The "arbitrary signals" might be approximated by normal distribtions with different sigmas, if this makes it significantly easier.

Answer 1

[Christoph F.]

I would try the plain vanilla approach first: Express the square wave as a linear combination of the given arbitrary signals, and solve the resulting equation system for the weighting coefficients.