How to set threshold for modulation recognition classification

[Shruti01]

Hello all,

I have generated digitally modulated signals such as ASK, FSK, PSK and I have done feature extraction for all these signals and saved those feature values in the database. Now when the tested signal is one of these signals, How do I set threshold value for these feature values to determine the feature value belongs to which modulation type.

Thanx in advance

 

[newtonoid]

What features are you using?

Compare histograms of the feature data for each modulation type for a large number of iterations/conditions and choose thresholds accordingly. Here's an example:



The above graph overlays two histograms. The green histogram represents the distribution of values of a specific feature of ASK signals for 400 or so algorithm iterations. The blue represents the distribution of values for PSK signals over the same number of iterations. With this data, for this specific feature, I could set a threshold around 300 -- above which I could assume ASK and below which I could assume PSK.

 

[newtonoid]

Shruti, are you still working on the modulation recognition? I currently am wrapping up a project on automated modulation classification and have a project worksite here: https://sites.google.com/site/thesdramcproject/home


The worksite is a little barren right now, but I plan on updating it with much more information on how our algorithm works and how it might be improved. Also, here is a quick demo of what the output of the current version looks like :smile::

https://www.youtube.com/watch?v=iFrWE6P-UNY