ML Algorithm or approach to solve timeseries selection

I am fairly new to ML and I’m working on a problem, but not sure which algorithm to choose. The dataset contains a set of incremental time-series events, in consistent and set intervals, with a list of potential observations. Something akin to what the output of an audio recognition program might produce where it looks something like:

I want to be able to have the program pick the best possible option based on previous values, so the output would like something like:

1 observation per time interval, with the best guess based on the previous values.

I did some research on RNN and LSTM, but those seem to be mostly used for predictions and forecasts on what the next interval probably is, not choosing from a set of captured observations.

Would anyone be able to point me in the right direction in terms of algorithms/approches to best solve this problem?

Any help is appreciated.