I am creating an algorithm that tries to estimate three time-series that should not cross each other (voice formants). The problem is sometimes the algorithm mistakes one time series with another and the algorithm "jumps" track. Here the blue line/orange line shares one track and orange line/green line also shares a track.
I was wondering if there is a name for algorithms that would fix these kinds of issue. An algorithm that estimates the best fit lines that minimize the number of sharp jumps for example. If this is a developed field I'd appreciate it if someone can point me to the right direction.