Is there an efficient and not complicated algorithm for searching for pronounced peaks in 2D arrays without ML?

Question

Algorithms

Tell me, for example, for such an array:

let arr = [
[4000.123, 0.0002250716, ],
[3999.641, 0.0002206956, ],
[3999.159, 0.0002314698, ],
[3998.677, 0.0002602464,],
//....... length 5678  
[2998.195, 0.0005059349,]
[2997.712, 0.0004620409, ]
..........
]

I tested different data, running it through simple conditions in reduses and loops, but it's not that. I get the data from the software connected to FTIR in the form of a csv file, but the software itself shows very clearly the main bursts. (exports in other formats - proprietary such as Perkin Elmer and so on, I did not find any libraries)
thanks if you put some common sense in this topic.

Answer 1

As far as I know, there is no strict algorithm for searching for peaks in an arbitrary time series. Therefore, one should be content with empiricism.
Well, that's what I would do.
1. Fix the size of the sliding window.
2. At the first position of the window, I would look for the maximum.
3. Move the window so that it starts from the found maximum point.
4. Again, I would look for the maximum.
If the maximum on the first window remained the maximum on the second window, I would consider it a "pronounced peak". If not, repeat the procedure.
The issue of choosing the size of the sliding window is solved empirically.

Answer 2

You can find the difference between neighboring elements. If the modulus of the difference is greater than some value - a pronounced peak.
The fixed value can be taken as the mean of the differences or the median or something else.