RF Signal Classifiers with TensorFlow

This is an SDR tool I wrote called 'Analyst'. It is used to select, view, extract and classify Radio Frequency signals. It has a built-in machine learning used for classifying known and unknown signal types. Signals can be viewed both in time and frequency domain with zoom controls. This allows the user to examine the details.

[@nickrhee7178]

Nice tool,. How did you compute the confidence numbers?

[@DimoDrill]

Nice work Steve! Can you please provide any papers to read about your method? I'm interesting in how do you prepare signal samples from waterfall for feature extraction. Do you process it just like a picture and use standard feature extraction metodics for 2d, or use some filttering and formalisation before to extract some features classic way (measure bandwith etc)

[@hellhelll1376]

Hi Steve. Is it possible to download Analyst somewhere? Really interesting software.

[@Dorrrr1111]

Hi Steve! looks amazing stuff, any idea to share the code for that? looking to use something similar to train TF for audio signal

[@HairfarmerSteve]

Unfortunately, I can't directly share code for this. What I can recommend is that you convert the timeseries data to a frequency series using an FFT, then build a training model with the small FFT segments as 1D inputs. The shape of the frequency components of the signal will be what your tensorflow algorithm can key off of.